 So thank you for this very kind introduction. Thanks to the president of the society also for having me over. It's a real pleasure. Let me just check whether people can hear me okay. First in the room, yeah, is that okay? Good, okay. And there's some sort of microphone in front of me which may or may not be relaying my voice to the Zoom audience. We'll hope that it is. And I guess people will send in messages if they can't hear me. So a real pleasure to be here. When it comes to memory, I have to say that I suspect the type of memory all of us are most interested in at the moment is immunological memory, the way in which a vaccine remembers something, a far more important form of memory right now than what I'm gonna be talking about. And I mentioned that because by way of preparation for today, I thought I might look at one or two YouTube videos of people talking about memory last night. But fortunately, my attention was distracted from such a task by the Richard Dimbleby lecture given by Sarah Gilbert from Oxford yesterday. And it was one of the most spectacular lectures I've ever heard in my life, how she relayed the amazing story and how it unfolded from one day to the next. And she remembered every little detail. It was fantastic to see an expert on vaccines, on immunological memory, remembering her own memories during the last amazing two years. It was fantastic. So what I want to do is to try to lay out to you how we make memories, how we store and keep them and then how sometimes things go wrong often later on in life. And try and touch on each of those themes if I can. And the central kind of argument is that we have within our brains a system which does kind of automatic encoding of information as things happen. And it binds that information to the context where we are possible also to the time where we are. And then somehow in a selective way it keeps some of this information and not others. And I'll try and lay out to you how the discoveries of that in which I played only the most minor part have unfolded over the last 30 years. But let's start with something we all know about namely remembrance. Remembrance are very important part of our culture. And when one thinks of the last 10 years what is the event perhaps that had the greatest effect on the country with respect to remembrance? I think it was remembering the 100th anniversary of the start of World War I in 2014 through to the end in 2018. And whatever one's feelings about the way in which remembrance sometimes gets too wrapped up in patriotic issues one has to recognize that many men and women lost their lives. And so I think we marked that and the whole of Europe in a sort of dignified way, remembering the loss of many lives. Now remembrance however is a somewhat unusual act of memory because particularly in the case of the First World War none of us were alive. So we're not remembering back to some event that was part of our lives, more a part of a kind of public semantics, part of our culture really. So that act of remembrance is generally of events or people that we may never have witnessed or be at events of great personal and public significance. So it's often more a case of learning new things than of recalling a past event. But it happens I live in Southwest Edinburgh right beside Craiglock at hospital which is shown as an image here. And that was the hospital which through the direct shift of WHR rivers treated the soldiers for shell shock. And that's where many soldiers went for treatment with a view to returning them to battle including of course the famous war poets Wilfred Owen and Siegfried Sassoon. And at the front entrance of Craiglock at hospital is a library and it has all of the books of these war poets. It's fascinating to read. And as I was looking through that I saw a book about Siegfried Sassoon that had been donated by a man called Dennis Silk. Now Dennis Silk was a school teacher of mine and Dennis Silk had befriended Siegfried Sassoon later in life. Siegfried Sassoon died in the late I think 1968. So it's just like I was sort of connecting in some way with my old school teacher and learning something new about the friendship between him and a soldier who'd been treated in the hospital just beside where I live. Now what we now do I think is often turn remembrance into something more of an art form. I think we all remember the ceramic poppies in the Tower of London. And likewise in Berlin they now remember the 10th, 20th and just this last year, the 30th anniversary of the fall of the Berlin Wall. And they do so with an element of celebration. I had the good fortune to be there on the 20th anniversary as part of a seminar and all the primary schools in the city had built bits of wall out of cardboard and painted them. And it was a way of kind of teaching the children about how important this was to the city. So that's the way in which sometimes we can turn remembrance into something that can be more joyous rather than a solemn. But of course, some aspects of remembrance are solemn. As you probably know, Angela Merkel stood down as the German Chancellor today. And she was at least German chemist. And so as she laid this candle at the memorial on the occasion of the 30th anniversary, one wonders what she was thinking. Because perhaps on the one hand she'd played such a big part in the reunification of Germany, but on the other hand, she may well have had friends who tried to cross the wall in an earlier time. Now, remembrance of course can be much more enjoyable. So I share one picture of my family and I marching in the march against Brexit in the summer of 2016. We had a great day. We lost that battle, but nonetheless, it was good fun surrounded by lots of other wonderful people. So what is memory? I mean, certainly in a school setting it's people trying to learn things of course. But most of the time it's incidental and automatic. Remembering what you had for breakfast where some recent event happened and what it was. But it can also be conscious and effortful, requiring a skill such as riding a bicycle for a kid or learning a foreign language. Here's Paul Newman in the film Butch Cassidy. And then there's also memory of the future, not just reminiscing about the past, but remembering to do something in the future. So there's lots of different dimensions to this thing that we all know and recognize as an important part of our lives. But what is it? If we tried to sort of pin down in some way, they don't want to get too bogged down by definitions, but nonetheless, it's some sort of change in response to experience in the broadest sense. It reflects our capacity and that of animals to change our understanding of the world and or our behavior in response to the things that happen. And on this view, it's a temporary or lasting change in the nervous system described technically as the creation of a memory trace or an engram in the brain. And then the expression of that memory changes our behavior. Now, moving on from that rather general definition, this I think implies that it isn't just a looking back as implied by remembrance. It's that learning, storage and recall are separate processes, components of memory that are necessarily intertwined. Processes through which we acquire distinct kinds of information, sometimes incidentally in coding. We represent it appropriately, if not always accurately through the processes of storage and then consolidation which happens during sleep. And then we express things later through recall, recognition or indeed these behavioral changes which I've been referring to and that's the retrieval process. So that's what memory is. And as you all know, it sometimes fails. Now, one of the facets of this automaticity is that sure here in a university setting people will be coming in to this lecture theater to learn difficult stuff in physics or in biochemistry or what have you. And they'll be paying attention to the lecturer and trying and their learning is of a very deliberative character. But that's not true for most of us during the day. We just do things and our attention moves around. And so it's not surprising that we forget where we put our glasses down because we'd be doing lots of other things at the time. So the attention may not be on the act of putting your glasses down and maybe on talking to your partner or your husband or your wife or what have you. And so there isn't attention paid to that action of putting the glasses down and hence you forget it. So forgetting is also a key part of memory and I'll come on to that in a little bit. Now, as Felicity kindly mentioned, I'm a neuroscientist. I'm interested in trying to understand what's happening in terms of the underlying physiology and the biochemistry that makes all this possible. And so here is the human brain, 100 billion cells, maybe 5,000 synaptic connections on many of those cells. And so tons of connections. Maybe those connections are part of what we need to think about, how you link one bit of information to another. Now, my next picture is actually my brain. And I thought since you've had the kindness to come out here this evening to listen to me, I thought I ought to at least prove to you that I do have a brain. So this is my evidence. It's a magnetic resonance image of my brain. And it was given to me when I was a subject in a human brain scanning experiment quite a long time ago. I also showed you because there's a great tradition in the Scottish universities that when somebody becomes a professor, they have to give an inaugural lecture. And that's certainly followed in the Faculty of Medicine in Edinburgh, which I'm a part of. And so I gave my inaugural lecture and I had the joy that my parents were both alive at that time and they came. And of course over the years, I'd given my parents lots of grounds for them to not believe that I had a brain through many of my misfortunes. And so I presented this image to the audience saying, you know, my mom and dad are here and I want to show them that I really do have a brain. At which point, my father, very loudly, such the entire audience could hear, said, my boy, that's no proof that it's actually working, he said. Not for the last time did he get one up on me. Anyway, let's go back to the real brain here. What I want to put to you is that what these cells form is networks which represent experience through their activity. And those networks can be quite local or they may be more distal. And we have this exciting new technique called diffusion-tensin imaging. And this is an image created by my colleague Mark Bastin in the University of Edinburgh which shows some of the major pathways between these different areas of the brain and its activity along those pathways that connects bits of information together. So things will come in through your sensory systems, these process perceptually, and then may want to link bits and pieces together using the processes of sadaptic plasticity. So let's make some memories, the encoding process. Then we're gonna try and keep them by storing in some biochemical or structural change in the nervous system such that you can then reactivate things later and then address further the question of forgetting. Now, we have lots of different memory systems in the brain, there isn't just one. I can't just point at that brain that I've shown you and say that's where the memory is. It's not like a chip in a computer in that sense. It's sort of localized in different places because there are different memory systems which have different kinds of jobs. And in particular, there's a sort of short-term memory which we're all familiar with like for remembering a telephone number for a short period of time or something of that kind. And that short-term memory system relies on the continuity of brain cells firing, continuous activity, which keeps the memory active. And as soon as you have interruption, you lose the information and you lose it forever. So this has a very limited capacity but it has astonishing fidelity. So that provided you're within the limit of the capacity of the system, then you can take those, say digits or letters or whatever it is and remember them very accurately. And our day-to-day experience of this now for most of us is two-factor authentication where we have to remember a six-digit number and write it into something quite quickly. And that's fairly easy. Most of us are able to do that, say in our banks and what have you. But the key point about this short-term memory system is the brain cells must keep going. And if they're interrupted, you lose the information. And of course, all of us can think of examples when somebody's told us the telephone number, there's been some distraction and then we have to be told it again. Now there've been detailed studies of selected neurological patients who have deficits in this system and that's helped to point the finger at the areas of the brain which are most involved in short-term memory and it turns out to be in the front. But what about long-term memory? Here, what you get is these structural biochemical changes in the connections from one area of the brain to another or locally within one brain area. And once those have happened, then the thing stays in a kind of dormant state for days, weeks, months or even longer. Such that then when a reminder queue comes in, it reactivates the set of cells that represent the information. But it's not that they're firing all the time in the way that happens in short-term memory. So if I were to say to you, can you remember back to what you were doing last Christmas? You can bring it to mind, perhaps, I guess, I hope so. But it's not that you've had to have that memory continuously for the last 12 months. Now long-term memory is subdivided into distinct components also, mediated by these different connections and in different brain areas. But long-term memory doesn't have the fidelity of short-term memory and it can change and it needs to update as we learn new things. So long-term memory is somewhat reconstructive. Now long-term memory is divided into two different types, procedural learning and declarative memory. And procedural is things like actions, like learning to ride a bicycle and things of that kind, where what you do is that you learn through multiple trials, trial after trial of trial, until you've got the balance right, you've got the steering right, got the pedaling right and so on. And that's what a child is doing as they're learning to ride a bicycle. Or sports skills like tennis or squash or what have you. Again, it comes through practice and numerous episodes of making mistakes, learning from those mistakes and getting things better. And that then develops into habits, which can then be run off without very much conscious awareness. You just sort of do them and you almost sort of find yourself having done them and not much more than that. But on the other hand, declarative memory is different and I'll come on to the various different components of that. But declarative memory compared to procedural memory is like the distinction, the philosophical distinction between knowing how and knowing that. Declarative is about knowing that something was the case. That a particular event happened, that I talked to this person yesterday, whereas knowing how is more like the skills. Now, a particularly famous patient called H.M. Henry Melasin, he's died now so we now are allowed to use his name, had very severe epilepsy in the 1950s. So severe that it was not being satisfactorily treated by the then available anticonvulsant drugs. So decision was taken in consultation with both him and his parents to excise the area of the brain where they thought these seizures were coming from, which is an area called the hippocampus. And this was done in an operation in 1954, long time ago. However, what they found after that unexpectedly was that H.M. now had no day-to-day memory at all. So lots and lots of tests were done on him, had the great privilege of meeting him before he died and I was introduced to him, talked with him for a while together with some friends. And then Suzanne Corkin suggested that I leave the room for a while, which I did, came back about 20 minutes later and he didn't know who I was and I had to be reintroduced. It was really quite striking to see the depth of his momentary amnesia. So here is a very famous figure called the Ray Ostroth figure and the original drawings shown at the top left. H.M. made a direct copy, which you can see is very good. It's got this little diamond thing hanging off. It's got the smiley face. It's got the railway tracks, et cetera, et cetera. He doesn't have a perceptual problem, but if you take it away and then 20 minutes later ask him to draw it, he doesn't even remember having seen it, let alone being able to draw it. So that would be an example of the narrative memory. But H.M. was also trained to move a stylus inside a star pattern, which you see at the bottom. But through a mirror, so that left and right were reversed, top and bottom were reversed. So it's quite difficult to do. All of us would have difficulty doing this. And what you can see is that on day one, he made 30 errors. He was constantly going over the lines. By day two, he was getting better. And by day three, he was hardly making any mistakes. He was developing this skill, this procedural skill, this habit, which he was doing without use of this declarative memory system, which had completely gone in him. So what happened on day four? Well, they brought him in and they said, let's do this mirror drawing thing, to which he said, what mirror drawing? Because he had no memory of having done it. And so then Suzanne said, well, let's try it anyway. He tried it anyway. And of course he did it almost perfectly. So this very, very clean dissociation between the procedural system, which is largely localized in an area called the Stray Atom, and the declarative system, which involves the hippocampus. Now, that dissociation is also shown sometimes in certain neurodegenerative diseases, such as Alzheimer's. And I thought I'd just show you now a little movie, which will give you a glimpse of that. An elderly man of 93 in a home with very, very severe dementia. But they discover that he used to play in a band, and so they bring in his old friends and he proceeds to go up to the piano and play with it. And it's very good. The staff at the home were blown away as we watched him play, and his wife started to cry as the memories came flooding back. So now there are further subdivisions. I told you that it was not a unitary system, episodic, semantic, and spatial. Now, whereas the procedural system requires many trials to get things right. We're constantly checking our errors. That's not true of the episodic system. Here is Seve Ballesteros, famous Spanish golfer who's died now, winning the 1984 Open in St. Andrews. And I didn't take this picture, but I was there at the time. And it was a magical moment. And for him too, his career, he won this fantastic tournament. He just walked up through the Valley of Death and putted his last putt and won the Open. Now, there's no trial too there. You don't have another girl. You do it in one go. But I suspect he remembered that event for the rest of his life because of the importance of the event. And I think that's the feature of this that we can remember back to a single thing. So there must be some other system which is mediating that. And the French photographer, Henri Cotier-Bresson who only died a few years ago at the age of 102, took these wonderful photographs of life in France, capturing individual moments which reflected aspects of family life or more public life. Here's a little boy who's obviously gone down the street to get some van-ordinaire for his granny. And you can see he's very proud of the task he's doing for the family. But we don't remember just events. We actually derive from that knowledge of events information that is some way abstracted from the events to become our knowledge of the world. And so here, what we call semantic memory is our factual knowledge of the world. And in this particular hierarchy, I could give other ones, we've got objects, we need to divide them into inanimate and animate. We then need to divide the animate ones into mammals and birds. The birds have to be divided into flying birds and flightless birds and so on. And so one builds up a kind of tree structure and that will come and be built in our brains through these multiple episodic events, also through formal training as well. And that component of declarative memory you can see is very different from the episodic system. Now, the third is spatial memory is a little Google map of Edinburgh. And we got leaf docks at the top and the city laid out around us. And of course, this is very familiar to me having lived in Edinburgh now for 30 years. And you'll have a similar map in your mind of the city of Glasgow. So we've got these sort of three components and what is it that's sort of putting all this together? Now my central argument then for this evening is that we remember events with respect to the context or place where they happen and often the time that they happen. And we do this automatically and incidentally. So if you had the misfortune today to have seen a car accident, hope you didn't, but suppose you had. And then you said to a friend, I've just seen a car accident. And they said, well, was anybody hurt? And you said, well, don't think so. Where did it happen? Wouldn't it be strange if you said to them, oh, I don't know where it happened, but it was definitely a car accident. That doesn't happen. That wasn't that you went out that morning trying to learn about car accidents. There's nothing deliberative about it. That type of memory just happens. It's a sort of incidental automatic thing. And it's, I think an automatic binding process by which the what, where, and the when get fused into a memory of an event or episode and a specific area of the brain called the hippocampus does this. Now, as I say, the brain later starts to recognize regularities, stripping away the context information and starts to build semantic frameworks called schemas. So we go back to the birds example I gave you a moment ago. If I said to you, what color are canaries? You probably say yellow. Then I say, well, when did you learn that? I take it that everybody in the room would say, God, I can't remember when I first learned that. You've lost all that extra information. You just derive from that to create a kind of semantic network. Now, let me just give a little story, which many of you will know, but it's a lovely story, which illustrates the power of this. And it comes from a lovely book by the late Francis Yates, who was a historian of the Middle Ages, went to the Cortle Institute in London. And this story, I apologize to those of you who know this story well, but it's nice and I like to show it. So I'll read it out because it may be difficult for the people on Zoom to read these things. So, Atta Banquet, given by a nobleman of Thessaly named Scopus, the poet Simonides of Seos chanted a lyric poem in honor of his host, but including a passage in praise of Castor and Pollux. But then Scopus meanly told the poet, he would only pay him half the sum agreed upon for the Panagiaric, and that he must obtain the balance from the twin gods to whom he devoted half the poem. Now, a little later in the banquet, a message was brought in to Simonides that two young men were waiting outside who wished to see him. So he rose from the banquet, went out, but could find nobody. During his absence, the roof of the banqueting hall fell in, crushing Scopus and the guest to death beneath the ruins. And the corpses were so mangled that the relatives who came to take them away for burial were unable to identify them. But Simonides remembered the places at which they'd been sitting at the table was therefore able to indicate to the relatives which were they did. So the invisible callers, Castor and Pollux, had hamsonly paid for their share in the Panagiaric by drawing Simonides away from the banquet just before the crash. And this experience suggested to him the principles of what came to be called the art of memory of which he's said to the inventor. Noting that it was through his memory of the places at which the guests have been sitting that he'd been able to identify the bodies. He realized that orderly arrangement is essential for good memory. And I think it's a really important principle in education, in our lives generally, far more important than some kind of magic pill to improve your memory, orderly arrangement. I believe, and I think there's now pretty overwhelming evidence that the human hippocampal formation shown in this diagram from David Amaral encodes events as they happen and binds them to the context. Now, how does it do that? Well, if we look at the anatomical connections which the hippocampus in the center of this diagram has, it receives information from large parts of the brain and in through a serigen called the enterinal cortex which degrades in Alzheimer's disease, but it receives the information about the context and about the what. So it binds the what and the where and then through connections up to the frontal cortex is linked in to the temporal information. Now, we've come to learn about that through some very important animal experiments that will be done over the years, particularly by John O'Keeffe working in London who discovered what he called place cells in the hippocampus in rats. Now, I know some of you may have some reservations about animal experiments, but they were extremely important in learning about the way in which the hippocampus sort of mapped space and he was able to record the firing of individual cells and some of those cells in that high proportion of the cells fired when the animal was in a particular place. They stopped firing when the animal moved out of that and other cells started firing. So these came to be known as place cells and you can see a little hotspot for a particular part of a circular arena. Now, soon after Edvard and my Brit Moser working in Norway, they had previously been post-docs in my lab and then they got jobs back in their native Norway and built up their own lab where they made amazing discoveries of what will come to be known as the grid cells. And the grid cells fire in a tessellated pattern which you can see in the diagram, providing a kind of metric of space and it's roughly speaking in a rodent about every meter in different species, different amounts. And another group of cells is the head direction cells which actually tell you which direction something is pointing in. So this information is all pouring into the hippocampus. Now you say, well, that's just in the rat. Well, true, but this system has now been shown in bats, lots of other mammalian species and even in humans too. Now human experiments are obviously rather difficult to do because not everybody wants to have electrical wires put into their brain, but there are sometimes occasions, particularly for the surgical management of epilepsy when this is done, in order to try to map out which areas of tissue are damaged and which are still okay. And a number of experiments like this have been done and the existence of place cells has now been confirmed in human hippocampus. Now this is not true for the grid cells for the enterinocortex, but using non-invasive functional imaging, a group both in London and in Germany have identified that some kind of grid cell like pattern can be observed using functional brain imaging. So I think we've got every reason to think that these discoveries from these animal experiments can be generalized across both a range of species and including humans. In 2014, John O'Keefe together with Edvard and Margaret Moser won the Nobel Prize of Physiology and Medicine for these key discoveries. And they had published in the run-up to that a number of really very impressive papers. And I had the privilege of going. It was a really fantastic occasion. Now what about other animals? Well, there are suggestions in the literature, which my two colleagues here know far more about than I do, but I'll just pass over this quickly. That other animals may also have this kind of episodic-like system of memory. And it's been some lovely work done by Nicola Clayton, both in California and now in Cambridge, which has also pointed to the existence such system in birds. But here are some work done by Serge Dahn of Grönchen University in Holland, where they've been tracking the movement of kestrels when they wake up in the morning from their roost and go searching for their breakfast, which is a little mouse running around in the field. So at the top there, the kestrel wakes up at 5.50 and flies in a little circle and then gets his prey at 6.15 in the morning. That was on August the 12th. And the next day, the kestrel wakes up at 5.43 or at least he flies off at 5.43 and goes to almost exactly the same spot where he was successful the previous day and gets some prey at 5.47. And then the next day flies off at 5.40 in the morning and is successful in getting prey at 5.52. Now this may be just a procedural habit, but there was an indication from the drift if one looked across further days that they were actually just remembering back to the day, the previous day where they were successful and going off, suggesting that they have something like a kind of episodic system which links the event of catching a prey to their knowledge of the entire environment. So I think this is a pretty general system. Now I developed a piece of apparatus in my laboratory called the water maze which is intended to try to test this notion from the physiology that this system existed in the hippocampus. This was published many years ago now, but I just thought I'd briefly show you this working so that you've sort of seen it with your own eyes as you may have heard about this. So what it is, just a large pool of water, there's a platform hidden underneath the surface of the water, it's a particular point. And we make the water cloudy initially with milk powder and more recently with material that we can get from builders supplies. So you see the animal swimming over and then climbing onto this platform. Now the important thing here conceptually is that the animal cannot see the platform. The platform's silent. We can swirl the water around such that it offers no particular smell. It can't touch the platform until it's already got there. So in effect, the only way that the animal could do this is through the existence of some kind of mapping system which would tell him the place where this platform might be. And since John O'Keeffe had indicated that this might be in the hippocampus, it seemed a worthwhile way of looking at things. So if I go on, I'll just check. Yeah. Now what we found was that if you damage the hippocampus, we got deficits of recent memory, not short-term memory, but recent memory in this situation, recent spendable. Also if you interrupted these input and output pathways that led to the hippocampus. And fortunately, others went on and replicated and did extensions. And so the observation I think is fairly sound. But then what I went on to do was to say, well, it's okay to damage the hippocampus, but how is the hippocampus doing this? And our supposition then was it was doing it by changing the connection strengths between neurons and in that way, storing information. And I discovered through using drugs that blocks anactic plasticity, that that would also cause an amnesia. Now this brings me to a very famous man, Santiago Ramon Cajal, who was the first person to sort of really draw at the level of individual cells and their connections, what might be existing in the brain. And these really remarkable drawings still exist. They're in the Institute of Kahal in Madrid. I still haven't been to see them, but I'm dying to see them at some point. But I have seen some replicas of them. They're truly amazing diagrams. And he was the first person to suggest that memory might be some change of the growth of these connections. So he was right. It took many years before the evidence for this was to materialize, but it is an idea that goes back a hundred years. It's not my idea in any sense at all. Now the first observation using physiology was made by Tim Bliss together with Tario Lomo in Oslo in Norway. And they showed that particular patterns of stimulation would actually strengthen connections. And as you can see, can make potentials larger here. This potential is larger here than it was initially. So this is evidence that these synapses can be made stronger using electrophysiological evidence. So my study was then to block that, which I did with a drug with a fancy name of DLAP-5, and indeed the animals failed to learn the water maze effectively. I'm sorry, had I been able to show you the film, it perhaps would have been a bit more convincing. So now let's do what Kahal would have wanted us to do to look at the tight intrinsic circuitry of the hippocampus to try to understand how this could actually work. And although this is just a Mickey Mouse diagram because we're dealing in the human brain with millions of cells, nonetheless, there's a very specific architecture of the dentate gyrus, which is the first bit of the hippocampus, then area CA3, which is the next bit, then area CA1, which is the next bit. And it looks as though what's happening is there's initially a kind of pattern separation system in the dentate gyrus, then area CA3 has some very nice re-entrant circuitry for learning sequences and area CA1 receives information from the entorhinal cortex about the context. So what it's doing here is somehow binding information about a vent which has sequences with context information. That's what this circuitry appears to be doing. So it's interesting, it's merely suggestive, but I think that it's, I think particularly interesting that the synapses can change in efficacy and Graham Collingridge was to show that there's a specific receptor, the NMDA receptor, which can change the strength of the connections at these synapses. So the drug I was using, AP5, blocks the NMDA receptor and it therefore stops the changes in strength of the connections. We're right down to a sort of molecular level and trying to link that back up to the cognitive and behavioral level. Now let's just go back to this. I've said we've got a pattern separation thing, then sequence learning, then event context binding and that's my suggestion that this is the sort of what, where, when system in the brain. Now what's striking is that if you look at contemporary deep learning algorithms, the kind that the company Google is using in many of its bits of software and there's this big institute now just near King's Cross station Google Deep Mind which has been developing a lot of very remarkable bits of software, one which is now called AlphaGo, which is one of the go championship of the world against professionals and so on. Then it turns out that the internal architecture of that software involves input layers, hidden layers, hidden layers and output layers and synaptic learning rules, sometimes more complex than I've described for the hippocampus, but this is inspired by this what's called brain inspired architecture of a form of learning. And another program which some of you may use a lot is Google Translate, where you can put a word in in one language and it'll give you back. That is just one of these sort of deep learning algorithms to actually in that case derive semantic knowledge rather than remember an event. But I think it's fascinating that a really detailed understanding of the anatomy of the brain has been an inspiration for people developing these networks. Now I'll move over that because time's running on but my basic point is that we can change the strength of these connections and have potentiation or we can weaken them with depression. Now what about keeping these members? I won't say a huge amount about this, but there are lots of things, attention, you've got to pay attention, reward and spacing, spacing information apart is good. You don't learn all in one go, space it out. Retrieval practice has been shown to be extremely important, particularly in language learning. Novelty around the time of memory encoding is something I've done a lot of work on and I think it's interesting and I'll just touch on an aspect of that and then prior knowledge and of course sleep-aiding memory consolidation. So if you like the initial encoding process creating what we call an eligibility trace, it doesn't guarantee that you'll get a long-term memory. You'll have encoded something, you'll got a representation, but then these other things will kick in to determine whether this information is kept. Now, I said that I've been particularly interested in Novelty around the time of memory encoding. Well, there's a very famous set of memories which are sometimes called flash-balled memories. The first of the fact is certain events are very well remembered by the particular generation for which they're important. I'm just old enough to remember the assassination of President Kennedy, I was only a young kid at the time, but I'm old enough to remember it. Assassination of John Lenn, 1980, Princess Diana's death. I remember exactly where I was. I was in Sacramento in California in a hotel and we were all watching the TV. 9-11, of course, and perhaps for younger people, the student killings in the Toya near Oslo, 2011. Now, if you take, say, 9-11, of course the images of the Twin Towers on fire and collapsing are horrific images and all of us can bring those to mind, but there's a further important feature is that we also remember many trivial events of that day. In the way that normally we forget the sort of trivial inconsequential things. We don't need to remember it. I mean, I guess we don't really need to remember all the events surrounding 9-11 on that day, but we do. So what's sort of going on there? Well, I've worked on that with some animal models and also with human experiments. And it looks as though what's happening is that when something really surprising happens, there's a release of neuromodulatory chemicals around the brain, particularly dopamine and noradrenaline. And these modulate the activity of the brain in such a way as to kind of give us a sort of print now, keep this information, it might be important. And we've been able to show that through, as I say, various kinds of experiments. And I think it's fascinating that it isn't just the novelty of the event. The Twin Towers, of course, was shocking, but it was also surprising. But that novelty creates a kind of penumbra which enables to retain other information. Now, these are the striking examples, but I suspect it happens quite a lot in our everyday life too. Things which are obviously less surprising than these examples, but nonetheless, if you have a surprise one day, you'll remember lots of features of the context of where that's happening. Now, the other thing is prior knowledge. I've been talking as though you make a memory first from the information coming in. Because you already know a lot. And what you know already influences what you're interested in, what you're gonna pay attention to, your ability to understand something, and so on. And so that is a further factor that we have to bring into the equation. That it isn't just the hippocampus teaching the rest of the brain, it's that the rest of the brain feeds down to the hippocampus prior knowledge which can influence the way we process things. Now, I'd love to give you some really genuine memory examples of this, but I think that a classic study remembering an experimental psychology may serve my purpose. So what I'm gonna do is I'm going to read to you a story and I'm gonna ask you all to imagine that you are subject to an experiment where after half an hour of doing some irrelevant activity like lots of multiplication problems or something of that sort, you've got to repeat of as much of the story as you can. Now, the story is grammatical, but you may find it a bit puzzling. Here we go. If the balloons pop, the sound would not be able to carry since everything would be away from the correct floor. The closed window would also prevent the sound from carrying since most buildings tend to be well insulated. Since the whole operation depends on a steady flow of electricity, a break in the middle of the wire would also cause problems. Of course, the fellow could shout, but the human voice is not loud enough to carry that far. An additional problem is the string could break on the instrument, then there could be no accompaniment to the message. It's clear, isn't it? That the best situation would involve less distance than there'd be fewer potential problems with face-to-face contact where the least number of things could get wrong. So that's the story and I then take you off and make you do lots of mathematical problems and then half an hour later ask you to write down the story. Now, some of you may be able to do it perfectly, but I hazard that a lot of you would find that quite difficult. Now, why? Why is it difficult? And it's not, as I say, it's a grammatical story. There's nothing grammatically wrong with it. It's that we lack information about what on earth is going on and it appears disjointed in some way. So what prior knowledge is sometimes doing is giving us this background information such that we can interpret new things. So I can mimic that by showing you a diagram. Here it is. So it's a Romeo and Julio scene in New York with a skyscraper and Romeo is playing the guitar tune, Juliet on the balcony. Right. If the balloons popped, the sound would not be able to carry since everything would be away from the correct floor. The closed window would also prevent the sound from carrying since most buildings tend to be well insulated. Since the whole operation depends on a steady flow of electricity, a break in the middle of the door also kills problems and I could go on. I hope I make my point that what the image I've given you is just a sort of experimental way of trying to show to you how prior knowledge can be so important in interpreting new information coming in. And that's very true in a university setting of course where we sequence our courses in such a way that the information acquired in first year is then used in second year, information acquired in second year becomes a kind of background for the third year and so on. So the progressive development of knowledge, particularly complicated bits of knowledge has to be done in a structured way because only then would it be possible to interpret things. And it's quite interesting from an educational point of view because there's been a move particularly in medical education towards problem-based learning. But now the pendulum's swinging back a bit not against problem-based learning, it's great to have to work, but you have to have a certain background knowledge. Otherwise the problem-based learning is just simply throwing your hands around without any detailed knowledge of what you want. So you need to have a balance of these two. Now this leads on to the idea that we have schemas. These were first proposed by Frederick Bartlett an English scientist from before the War Jean Piaget and John Bransford who developed stories like the balloon story. And the notion then is that these schemas provide a kind of framework into which we can abstract information when we transform something from this episodic event thing into a semantic state. And I've done experiments jointly with Ian Fernandez in Holland using human subjects and functional magnetic resonance imaging and have examined patterns of activation in the brain when the process of knowledge assimilation is actually going on. And I think I have one image of that so that here on the right in panel B we see the hippocampus being activated, that's important. But what's also important is the activation in the cortex at the same time. That's the prior knowledge that is helping to assimilate the new information into the established structures. I just finally touch on the losing of memories. I said earlier on, but not remembering where you put your glasses is just sort of trivial. It doesn't really matter too much. And of course that's true. But it's quite interesting. Dan Schachter who's a very distinguished experimental psychologist at Harvard wrote a book called The Seven Sins of Memory. And one of his sins was transins which is just the weakening of memory traces. The sin of absent-mindedness, breakdown of attention. Clearly if you don't attend to things, this automatic encoding might happen but if you haven't attended properly you're not gonna perform a proper representation. Sin of blocking has to do more with retrieval. You've got thwarted memory research. You can't find the thing you're after. The sin of misattribution, assigning to the wrong source. Telling a joke back to the person who told it to you yesterday, that kind of sort of thing. Suggestibility. This is very important in a legal context. The idea of implanting information by means of leading questions. Bias, editing and rewriting. Something that I think we're all subject to the sin of persistence. And what Chakta says is all of these failures are exacerbated by age but that they're not really failures at all as they reflect the proper ration of a finely tuned system, not vices, but virtues. He goes on, the seven sins are byproducts of otherwise adaptive features of memory. A price we pay for processes and functions that serve us well in many respects. So I guess my message here is don't worry too much about occasional forgetting. Your memory system is probably working pretty well. And intelligent errors occur in a number of cognitive systems, including memory, and they reflect features of systems that are normally relatively faultless. Let's take the sin of persistence. Clearly, something like post-traumatic stress disorder can be very disabling. And one would like to be able to do something to relieve people from these flashbacks to some horrible accident or event in their past. But you've got to look at it from an evolutionary point of view as well, which is that the persistence memory of successful escape from any life threatening situation is, of course, adaptive. And you can't have both, is it? I mean, you do get both, excuse me. So if you've got a system which is enabling you to remember things that really, really matter, then when you have something frightening that really, really matters, you remember it. The sin of transience, forgetting things quickly is annoying and we fail to pay attention, but imagine what it would be like if we remembered everything. Now, when I go back to this synaptic plasticity, there are far more details I could be sharing with you, but one very interesting feature of long-term potentiation, which Mr. Loma had discovered, in which we worked on when looking at synaptic connectivity in the hippocampus, was that most forms of LTP decay down to baseline and they parallel are normal for getting during the day. But that when these novel events happen or when other surprising things happen, then protein synthesis kicks in and stabilizes synaptic change such that what you're left with at the end of the day is just a subset of all of the things you've encoded automatically during the day. And that subset is then subject to consolidation during sleep overnight. So the system's an example of the way in which it's finally balanced. However, sadly, as I think we all know, memory and other cognitive functions do begin to fail in certain degenerative neurological diseases that deter in older age. And here from a poster from Alzheimer's Scotland, example of somebody's put the Sardine tin in the medicine cap. Now Alzheimer's disease is the most common form of dementia. It now affects 55 million people worldwide. It's occurring in all races and ethnic groups and it's extremely expensive in terms of the health budget and also disruption of the life of carers. People are working on effective treatments but it's a slow process. And in many respects, we've got to recognize that dementia in some ways affects the carers as much as the victim. Revealed, I think quite nicely in John Bailey's account of his wife, Iris Murdoch during loss two years of her life. Now, one thing about the dementias is there's a forgetful early stage which is insidious and gradual. Particularly elderly relative starts to be more forgetful than they used to be in the past. And then in addition, time and space disorientation happens as well. And we now know in that early stage, cells in the enterinal cortex which are projecting to the hippocampus start to go. Then there's a confusion in the middle stage with very obvious memory deficits and perhaps the need for supervision. And at that point, the patient or the person afflicted starts to have quite a lot of problems of anxiety and they may get very angry as well with good reason. Then in severe dementia, there's more widespread cognitive impairment, marked personality and behavioral problems, quite a lot of epileptic seizures that's perhaps not very well known about, but it does happen. And then there's a terminal final stage as well. Now, amyloid plaques as shown here and tangles these black things are the sort of hallmark neuropathological lesions of Alzheimer's disease and are seen in the brains from such patients in post-mortem histology. And they seem to arise from various causes but have been studied quite carefully in people who get dementia of a familiar kind, i.e. a genetic form. It's a relatively minor group who do that but they've given us insights into the pathology. So your cause might be some sort of gene deficit, giving an accumulation of a particular peptide called amyloid beta. And what we're not really sure is what happens then creates these plaques as I've shown you, but does it damage the cells? Does it cause neuronal dysfunction? Then there's something else called tau, which also happens, which is part of the tangles. And here you can see these neurofibrillary tangles. So one of the real problems in this research has been that if you just start with post-mortem histology, you're looking only at people at the terminal stage and you've got really very little insight into the sequences of things. However, once it was realized through the work of John Hardy at University College London that there were mutations of amyloid precursor protein, the precursor protein of amyloid beta, such that there's the 670, which is Swedish mutation, 717 is the London mutation. These mutations result in aberrant processing of this A beta molecule. Now, once we knew that it was then possible to do a particular type of animal experiment, namely to engineer into mice, as it happened in these experiments, these very aberrant mutations and see what happened to the animals. Let's just pause for a moment to think about that experiment. In one sense, it's a great idea to do it. On the other hand, if people are not getting this illness until they're 55, 60, 70, how on earth can a mouse experiment help us? Well, it has been done and a mouse has been developed. And I was had the good fortune to work with a company, Alain, which is an Irish company that research laboratories in South San Francisco and show that these mice became forgetful over time. Now, time is almost up, so I'll just race on. But Alain went on to develop an immunotherapy approach and they were able to show that in these mice who develop these plaques, as you see in the control, they could completely clear this through a vaccine. Now, the company was unsuccessful for other reasons, but recently just published in Nature a few years ago, there's an antibody called aducanumab, which has been developed, which reduces a beta plaques in Alzheimer's disease. So I think there's hope that we may be getting towards a treatment and this has now got FDA approval. So to conclude, what I've tried to do is to give you a little bit of insight into the way in which neuroscientists think about memory now. There's an encoding process, particularly of information into distinct memory systems in the brain and the crucial role of synaptic plasticity in just triggering distributed associative storage of the kind that you have in these deep learning algorithms. Numerous factors trigger retention, better attention, peri-event novelty, sleep, but also prior knowledge. And with respect to losing memories, forgetting doesn't always imply damage to our memory systems. And as Felicity said at the beginning, working on this is one of the grand challenges of our life. Now, just as a kind of little coder, and I'll take a minute, many of us now enjoy many facets of the externalization of memory through our smartphones. Now there's nothing really new in this. I mean, it goes back to Gutenberg and you mentioned a printing, when people didn't have to rely on just word of mouth, you actually had documents that then could be printed in people's native languages and not just in Latin and read in the monasteries. And so that memory became more dispersed through these things. But now of course, we have these devices such that a great deal of human knowledge is just to sort of mouse click away. And I think we all know this is radically changing lives of many teenagers. And a lot of good aspects of that, my son has learned a huge amount of both English and Spanish through his iPad. I'm not rejecting to that. So that's great. But one wonders whether this might be providing crutches such that we don't need our memory system. I think that's wrong. And part of the reason why it's wrong is because of my argument that so much of memory is automatic. We wouldn't want to constantly look at our phone to find out which cabinet the salt is contained in in our kitchen or something. It's more than just to have it in our brain to know it. Now there are circumstances such as GPS systems which are incredibly valuable. And I wouldn't be without mine, particularly when I'm traveling in a foreign city where I would get lost very, very quickly. But one wonders how far this will go and whether we will become completely dependent on these digital systems. And there's a lovely cartoon about this, about John Smith who on his gravestone put his email address so that he could stay in touch with everybody. So thank you very much. I'd like to express my enormous debt to John O'Keefe and my former postdocs, my Briton Edvard who have been such an inspiration to me with their physiological work. And if I think laid the foundations for the kind of framework that I've given to you and also to my family which has been fabulously supportive, Monica and the family's photographer, Adrienne. Thank you very much. I'm sorry, I'm not doing what I'm telling everyone else to do. So Paul asks whether the strengthening of neural connections in our brain happens in a way that's similar to artificial neural network algorithms, are there any training or are the training processes involved in the two completely different? Yes, I drew attention in the lecture to the similarity between the architecture of the hippocampus and certain deep learning networks. And the question relates to that. Well, I mean, yes is the answer but bear in mind that deep learning networks have a whole variety of different algorithms that they use for different purposes. And some of them involved a procedure that's called back propagation in which they feedback errors. And that's particularly important for these things which are more procedural nature where you've got to fine-tune something and it has to happen multiple times to be able to get to perfection. And those are used a lot in these chess and go games and so on. But a system just for remembering a particular image will work using the standard heavy and one-shot method and that's true in the machines as well. Okay, thank you. One more from the Zoom and then we'll take some questions from before. Why does dementia seem to enhance old memories but to jumble up newer ones? Ah, fascinating question. Fascinating question. What's mine? Who asked that? An anonymous attendee. An anonymous attendee. Well, I don't know that it's just dementia, isn't it? I mean, I think for many of us, remembering back 20 or 30 years sometimes is easier than remembering back just a week or so. And I think that part of the issue though of this getting accentuated in dementia is because if the system is starting to break down, you'll not be making new memories very quickly. But the older ones will have been consolidated over many, many years. And so provided the retrieval systems working very well, then during this insidious decline, you'll get a shift more towards earlier stuff. And this is actually being used therapeutically with music where it's clear that dementia pages find it very enjoyable to listen to music from the 50s or 60s or even earlier times and don't really have much interested in the more recent stuff. And it brings memories back probably to them by virtue of hearing these lovely old songs. Actually, there's one more question on this one which relates to that. Yes. And it's a very poignant question. Is there value in actively seeking to know if one has some form of dementia when there's nothing that can be done about it at present? It is a poignant question. So the question has to do with whether one should sort of take active steps to find out where one is on something like this. I mean, one can certainly find out whether you've got some sort of genetic abnormality but whether you would want to know or not, I don't know. I feel that a slightly divided view about the treatment for Alzheimer's disease which I'd share with you. And this may shock you as a suggestion but I think it's worth thinking about which is that if we could develop treatments which would enable us to retain our cognitive capacity, retain our memory for like six months or a year or something like that and having got the diagnosis, you could take that treatment and function effectively with your children, with your family, work things out. There are any things that need to be settled, get that done and still be able to do that cognitively. That I think would be enormously valuable. But once you get to later stages, my worry is that the treatments may just prolong the agony for everybody rather than do much more than that. So I sort of, well, if it happens to me, I'd take these treatments and then I'd hope that I'd have a catastrophic decline at that point. I mean, that would be my attitude to the thing. Right. Thank you. There's a question from the back. Can you speak, can you stand up and speak as loud as you can? Sure. I'll take your mask off if you don't mind. If I need. That's right. Yeah. Yeah, yeah. I need to just repeat. Yes, no, no, no. So the question is whether we were hearing about how particular experiences, particular biochemical changes in the brain can potentiate memory. And I think if I've understand the question rightly, it's can those be used, could we draw on those to help with the case of dementia? Yes. I think it's a really interesting idea. It has been an approach that people have thought about. So with respect to the way the glutamatergic synapse, which makes the excited to transmitter of the brain, the way those synapses work with these two different sets of receptors, one for changing the strength of the connections and one for expressing them, is there have been drugs which developed, which instead of blocking the NMDA receptors I did with the AP5 drug, actually try to make the NMDA receptor work better. And you can do that with glycine, for example. Now it's fine, but unfortunately glycine does different things than the spinal cord. So you can't just simply take a pill because you'll have side effects which are not very appropriate. So animal experiments you've done where you just direct the drug into the hippocampus and then you do see positive effects. But how you then translate that into a therapy where people could take a pill in the way we would take paracetamol, I think that hasn't yet been worked out. But I think it is an interesting idea, particularly as the system starts to sort of break down. Can we get a little bit more out of the system before it completely collapses? Thank you, we've got another question. Could you shout? I need to repeat that question. How do we know that memories are consolidated in sleep and that it isn't just dreaming? That it isn't just dreaming. Well, there have been lots of sleep laboratories that have been set up where you do all sorts of different kinds of experiments, beginning typically with electroencephalographic recording from the skull of the brain to see the different stages of sleep. So slow wave sleep goes through four stages and the fourth is the deepest one and then you go into REM sleep which is the one where your eyes move around, rapid eye movement sleep and that's when the dreaming typically happens. So there've been lots of discoveries about these different stages and the particular waveforms that you see in the brain. But on top of that, people have also been trying to do experiments of the following kind which is you would give somebody some task to learn, then let them sleep and then test them later to see if there's some beneficial effect of sleep. Other subjects might be sleep deprived and do they show no benefit or even get worse? But the more you think about these experiments, there's so many confounding factors. It's often very hard to work out. Are they doing badly in the test the next day just because they're tired or is it because they didn't have sleep consolidation of the memory? So people have tried then timing things at different times a day where you do this nine o'clock in the morning then you test them 10 hours later or 10 o'clock at night then you test them 10 hours later and then you've got sleep in one case and not in the other but then you have circadian influences on this. So it's actually incredibly difficult to design what you might regard as a sort of definitive and adequate experiment. Now, one of the things that has happened is that people have noticed that there are particular spindles called cortical spindles that appear to be connected as discovered through animal experiments with something called sharp wave ripples in the hippocampus. So you get a sharp wave ripple in the hippocampus and you get a cortical spindle. Now in the animal experiments what you can show is these sharp wave ripples actually contain information. So if you've got an animal which is visited various place cells and place cells are fired you can see those place cells firing during these ripples. So it's as if information is being sent for assimilation into cortical networks but there's still a lot of uncertainty in this field because it's very difficult to do these experiments and the animal experiments I think are very inadequate in various ways. I mean, how do you know about dreaming in the laboratory animals? It's very difficult to look at. That's a little bit of a question for one. So the question is is the capacity of the brain to store information limited? Can our brain simply get too full? I don't think the cortex gets too full. I think there's no reason why you would be using up the amazing capacity of the cortex for long-term storage. But I think hippocampus gets too full. Gets too full every day. And this is also part of a wider aspect of this theory that if you have an automatic system it's encoding information. And by the end of the day it's pretty full. Then stuff starts to decay you pick out the important stuff and then this cortical and then this sleep-associated consolidation keeps the stuff that matters. But you then have to clear out or a decay process has to clear out the hippocampus to get you going for the next day. So I think the hippocampus does get too full but the rest of the brain probably not. Probably up from the floor and then what's about to be done? Sure. That's okay. That's fine. I think you're about to get out there. Yeah. Right. Exactly. I think it's just... Sorry. So the question is is there such a thing as muscle memory say like a pianist might have for playing a tune sort of automatically without too much conscious attention? I think it's just a phrase. I think so. I mean, maybe musicians here who know more about this but I think that the sequences will be encoded in the motor cortex a stripper cortex in the top here. And there are cells there called bet cells which have long fibres down to the ventral horn cells which then control the fingers and the arm movements. And so the sequential information or being able to play a tune that the pianist knows very well it's up here. Now that doesn't mean to say that there won't be muscular aspects of the way they hold their hands the way they sit their body and so on. All that has to be done right, everything. Otherwise, they won't be able to do it without paying attention. So, but I think muscle memory probably is just a loose term. Just read yourself. Sure, sure. It's a comment and a question. Right. This is from Gordon Barkley who's been listening on Zoom a psychiatrist and trauma therapist. I'm interested in your comments on what to many is the very important and practically relevant theory of memory reconciliation in the context of transforming trauma memories. And he says thank you, great talk. So perhaps that's a good point to bring this session to a close and thank Richard for a great talk and you could go on questioning you for hours but we'll stop. Okay, thank you very much.