 inhabitants are not behaving erratic, random. They exhibit precise behavior. My guess is that precision marks Professor Candale's teaching. I'm sure that enthusiasm and joy and abounding curiosity also mark his teaching. He teaches in the College of Physicians and Surgeons at Columbia University. Graduate students and postdoctoral students there do research under his guidance. He has transmitted many messages to many students, and those students join him in sending messages to all of us who are interested in what happens in our brains. Not only does Professor Candale study learning and memory in the brains of snails and mice, he has spent a wonderfully active and productive life committed himself to learning and to teaching. I look forward to being your student this morning, Professor Candale, as you talk to us about genes, synapses, and memory. Professor Eric Candale. Thank you, Professor Owen, for those gracious remarks. I also want to thank you and April Holzer for hosting me in such a gracious way during my visit to this college. In thinking about yesterday's talks by David Hubel and Tony DiMascio and Pat Churchill, I was really struck, as I think all of you must have been, by what an extraordinary period of science we are living through. And I would dare say that when historians of culture look back upon the last half of the 20th century, they're likely to comment that the most remarkable contributions to our culture life during this 50-year period has come not from literature or from art or even from philosophy. It has not come from psychoanalysis or the movies. It has come from the sciences. And I would argue, in particular, for the biological sciences. And this is true not only because of the great insights that have been achieved, but because during the last 50 years and particularly during the last 15 years, there has occurred two rather dramatic unifications within the biological sciences. The first, and perhaps most immediate, has come in molecular biology. The ability to sequence genes, the work of Wally Gilbert and Fettsanger, and to use that sequence to infer the amino acid sequence of proteins that they encode has revealed completely unanticipated relationships between proteins encountered in different contexts. As a result, molecular biology has been able to establish a plan for how the cells in different parts of the body of a single organism and different organisms, how they share a common plan. And this conceptual framework for the cell has provided a unifying framework for several disciplines in biology that were previously quite unrelated. Genetics, biochemistry, immunology, development, cell biology, and even aspects of signaling and neurobiology. As you heard in the three lectures yesterday, a second less dramatic because it's earlier in its evolution, but in the long one, I think even more profound unification is occurring between neuroscience, the science of the brain, and cognitive psychology, the science of the mind. This unification has provided a new framework for examining memory, perception, language, and one hopes as Pat Churchill indicated in the long range, selective attention and aspects of conscious awareness. This of course raises the question, to what degree can these two independent and in many ways disparate strands be united? Can molecular biology also enlighten the study of cognitive processes? In my talk this morning, I would like to outline for you the possibility of a molecular biology of cognition. I'd like to use several simple examples from different forms of learning in snails and in flies and even in the mammalian brain, using genetically modified animals and manipulations of genes in both invertebrates and vertebrates to illustrate that one can begin to understand one tiny component of memory storage in some detail by combining the techniques of cognitive psychology with those of molecular biology. As a result, the study of learning may provide the first molecular insights into a mental process and thereby join the circle of unification by bridging cognitive psychology to molecular biology. Let me begin by saying something about learning. Learning, as you know, is the process whereby we acquire new information about the world and memory is the process whereby we hold on, we store that information over time. Most of what we know about the world and its civilization we have learned. That's why you come to this college. As a result, learning and memory are central to our sense of individuality, our sense of ourselves. Conversely, as we often learn tragically, the loss of memory leads the contact with one's immediate self, one's life history, one's interaction with other human beings. Thus, this orders of learning and disturbances of memory point to developing infant as well as the maturing adult. Down syndrome, head trauma, mental retardation, dyslexia, the normal weakening of memory with age, the devastations of Alzheimer's disease, of Huntington's disease, these only represent the most common of a large number of diseases that affect memory. What then can the joint effort of cognitive psychology and molecular biology bring to the study of learning and memory? I think it's fair to say that the initial insights have come from cognitive psychology. The most dramatic insight has come from the realization which is emerged from the study of patients with memory disturbance is that learning and memory are not unitary faculty of mind. We do not use a single strategy to inquire information about different things in our environment. We use at least two different strategies, as was first shown in the work by Brenda Milner and has subsequently been shown by Larry Squire, Dan Schachter, and a number of others. These are commonly divided into explicit learning and implicit learning, and I wanna say a word about each of them. Explicit learning is the sort of learning you conventionally think of. It's learning about people, places, and things, and it has several very characteristic features. Explicit forms of learning involves conscious participation, so if you think about a person, for example, I think of Dr. Owen, it involves a conscious recall of what Dr. Owen looks like in recognizing him as he sits in the audience. This involves, as I will return to later on, a specific set of regions in the brain located in your temporal lobe, deep in the temporal lobe, involving prominently a structure that I will refer to later on called the hippocampus. Explicit forms of learning is sometimes subdivided into two categories, episodic, the historical events of your life, and semantic, what you learn in school. Now the amazing thing about explicit knowledge and explicit learning is that you can define patients that have a lesion in the temporal lobe that can no longer store new explicit information. These patients can learn to recognize Dr. Owen, but if they turn their face away and come back to him one or two minutes later, they will not recognize him. They will remember things that they encountered in people they encountered prior to the trauma to the temporal lobe, but after having a bilateral lesion, lesion on both sides of the brain, they will not be able to put new information into long-term memory about people, places, and things. But the amazing thing about those patients, which Brenda Milner first discovered, is that even though they cannot remember the name and the face of Dr. Owen, these patients still have access to a vast variety of information which they can store. That sort of knowledge is called implicit and it is remarkable in two ways. One is that it is completely unconscious, as I will tell you in a moment, and two, that it involves learning not about people, places, and things, but it involves learning about habits and motor and perceptual strategies. For example, when you hit a backhand and you rush to the net to hit a volley, these are moves that you make that you may initially acquire by conscious participation, but after you've mastered them, you carry them out without in any way thinking about the details, driving, riding a bicycle. These are motor skills and perceptual skills that you carry out without thinking about the detailed muscles that are involved. Brenda Milner discovered this when she was testing one particular patient by the name of H.M. that had a bilateral temporal oblusion. Even though she had tested it for years, every time she walked into the room, H.M. could not recognize Brenda Milner. She had him one day, however, do a task which involved motor and perceptual skills, learning to trace the outlines of a star while looking only in the mirror and not at his hand or the piece of paper in which he was drawn. Amazingly, H.M. performed like you and I would, made a number of errors the first day, but over four days became progressively better so that in a fourth day, performed exactly like a normal subject. In fact, it showed the same improvement that a normal subject would every one of those four days. Well, when she asked them on the fourth day, on Thursday, H.M., why are you doing so much better today than on the first day? H.M. said to her, what are you talking about? I've never done this task before in my life. This set of learnings which involves things like classical conditioning, operating conditioning, habituation, sensitization, very simple forms of learning, actually many of which were really studied extensively in the Midwest, which is really one of the pioneer areas for studying elementary forms of learning. Do not involve the temporal lobe, but they involve the specific sensory and motor systems that are recruited for the task. They also involve skills and habits of various sorts in a fascinating phenomenon called priming I will not have a chance to tell you about. Now, here we have two dramatically different forms of learning, dramatic different forms of memory storage involving completely different brain systems. And one would like to know is, when one looks at the cellular molecular mechanisms underlying them, is there a common grammar that both of these forms of learning share? And this is essentially the question I wanna discuss with you this morning. I wanna first consider mechanisms of storing long-term memory for implicit forms of learning by looking at a plesia and drosophila, and then I wanna focus on explicit forms of learning by focusing on the mammalian brain. Since this involves the hippocampus, which is a structure which evolved with vertebrates in this particularly well-developed in mammals, explicit forms of learning really is best studied in higher animals such as mammals. But implicit forms of learning, simple modifications of reflex behavior, perceptual skills can be studied in invertebrates as well, and people have studied them successfully in a plesia as well as drosophila. So I wanna look at the relationship of memory storage in invertebrates, invertebrates, looking at these two major classes of learning. But I wanna focus not on all aspects of memory processes, but on the particular components that both implicit and explicit memory share, which is the switch whereby a short-term memory, a memory that lasts minutes to at most hours, is turned into a long-term memory, a memory that lasts days and weeks. And before I get into that, I wanna, with some broad brushstrokes, paint for you the intellectual history of memory research and focus particularly on the discovery that there is a short-term memory process and a long-term memory process, and that they're related to each other in rather precise ways. We owe to Hermann Ebbinghaus, the German philosopher, the first experimental approach to memory storage. He actually carried out experiments in human beings, many of the experiments he carried out on himself. This is quite a remarkable accomplishment. Psychology, particularly in its early days, as you can imagine, was a very contentious field. There were lots of arguments. Yet despite the fact that some of these arguments continue, there is absolute agreement on Ebbinghaus' work, even though you would think he might introduce a bias by using himself as a subject. What he did was, he used word lists. Actually, they were extremely clever. Once he created a new language, nonsense words, so they were completely alien words like Esperanto. And he taught himself lists of these words, and he tested himself to see how he would remember it. And by this method, using this method, he was able to outline the major principles of memory storage. There were three. First, he pointed out that memory is not unitary. There's a short-term form, as I indicated to you before, the last minutes, a long-term form, the last days, weeks, and even the lifetime of the individual. Number one. Number two, in order to move from short-term to long-term memory, you often, not invariably, but often, need repetition. This, of course, is obvious to you. This is what your mother taught you. Practice makes perfect. When you repeat something, you remember it for a long time. And third, he and his students, Miller and Pilzegger, first pointed out that doing the transition from short-term memory to long-term memory, the memory process is highly susceptible to interruption. This is how this third point, the susceptibility to disruption was discovered. Miller and Pilzegger had subjects learn a list of words, these nonsense words, had them learn it repeatedly so that they remembered the list perfectly well when tested 24 hours later. He then took another group of subjects and had them learn the same list and tested them 24 hours later. But in addition, with the second group of subjects, he had them learn a second list, various times after they'd learned the first list. If he had them learn the second list, 10, 15 hours after they learned the first list, then learning the second list did not interfere with the retention, the recall of the first list. But if they learned the second list within the first hour or two of learning the first list, then learning the second list interfered with the recall of the first list. I hope I said that correctly so that you can follow in detail. But the idea is that when you learn a list, you really need the opportunity to consolidate that information. That's why that period is called the consolidation period. If you learn something very different, right after you've learned the initial information, that second learning event could interfere with the successful storage of the first event. This is of course why we have breaks between classes that is logically based upon the Ebbinghaus experiments. We obviously would like them to be longer and many people regret that Ebbinghaus didn't find a five to 10 hour consolidation period. Now, when this finding first emerged in around 1900, it got the clinicians all excited. Now, as David Schubert knows very well, it doesn't take very much to get clinicians very excited. They got excited because they had encountered this sort of consolidation phase in their clinical practice. They had seen patients who had convulsions, epileptic patients, and when a patient has convulsions, they often describe to you when they come out of the seizure that they have an amnesia, a retrograde amnesia for the period, often an hour or two, just before the convulsion. As if information was in short-term storage and about to be processed for long-term storage, but in this sensitive period, it became disrupted by the convulsive episode. This made people realize that this is in fact a very general process and people began to look for it in experimental animals. And in the 1950s, a psychologist by the name of Duncan actually took a group of rats and taught them a task that they could retain perfectly well for 24 hours or more. And then he gave them an electroconvulsive seizure, a single seizure at various times after they'd learned something. And he replicated essentially this finding. If he gave them the seizure five or 10 hours after they'd learned something, then the seizure did not interfere with their remembering it 24 hours later or subsequently. But if the seizure occurred within the first hour of learning something in this magic consolidation period, then the seizure interfered with the retention process. In the 1970s, a major insight into this consolidation process occurred. A number of different laboratories, beginning with the flexence at the University of Pennsylvania, but also Agronoff at the University of Michigan and Sam Barandes, who was at that time at Albert Einstein, who is now at San Francisco, found that inhibitors of protein synthesis selectively interfered with long-term memory without interfering with learning or short-term memory. Here is a typical experiment. Two groups of rats learning a particular spatial task. This is an explicit task. They're learning about where they're located in space. One group has received an injection whereby protein synthesis is inhibited. This is cyclohexamide. And a second group of rats receives just a saline injection. Both of them with repeated performance improve in the task. Practice is making progress. When they're tested for short-term memory, a quarter of an hour later, there's very little difference between them. The difference when it's found is not statistically significant. And in many experiments, there's no difference whatsoever. So short-term memory is minimally or not at all effective. But if you look at long-term memory, memory after three hours, you see that the animals treated with inhibitors of protein synthesis are back at basal level. They've learned nothing. They remember nothing at all of what they've learned. And the saline controls remember everything perfectly well. This shows that new protein synthesis is not important for learning. It's not important for short-term memory, but it's somehow important for the storage of long-term memory. Now, I need to emphasize one thing for those of you who don't think about protein synthesis every day, which is I hope most of you. This experiment should not be interpreted to indicate that proteins are not important for memory storage in the long-term. They should not be indicated to mean that protein synthesis is not important for learning and for short-term memory. Proteins are important for absolutely every thought you have, every feeling that you have. But the proteins that you use for learning, for short-term memory, for many things that you carry out in your life are proteins that live a long time. They hang around for a long period of time. So inhibiting the synthesis of new proteins for a period of hour or two does not interfere with the stability of those other proteins. These inhibitor experiments indicate that proteins that turn over very rapidly, that are activated or that have a very short lifespan, those are the ones that are important. A specific subclass of proteins are important for the transition from short-term memory to long-term memory. And this is shown most dramatically in the last of these series of experiments, which is a remarkable experiment. This is an experiment that repeats the same thing that happened before. Animals are taught a task so that they perform it perfectly well 24 hours later. This is actually a group of goldfish, I think I've lost my point to number one, a group of goldfish that have learned a particular task that 24 hours later they perform it perfectly well. Then different groups of these goldfish are giving injections of inhibitive protein synthesis. If they're given the injection three hours later, it has no effect on their retention 24 hours later. So if they learn the task and they get the inhibitor three hours after they've learned the task, inhibiting protein synthesis for a few hours does not interfere with their retention of the task 24 hours later, their long-term memory is perfectly intact. But if the inhibitor is given immediately after learning, and for the first hour or so, it interferes quite dramatically with their retention 24 hours later. So that suggests that during and right after learning, something dramatic is occurring in terms of proteins that turn over later. It looks very much like specific genes might be turned on that produce proteins that are essential for the learning process. Well this finding clearly caught my attention. It caught my attention for two reasons. One is here is a feature that was shown to be shared by both explicit and implicit forms of learning. Two, this is a feature that suggested that protein synthesis is important not doing the stable phase of memory, but in the switch between short-term and long-term doing this consolidation phase. And three, it suggested that this switch might actually involve a gene switch whereby specific genes were being turned on or off. And if this was so, then one could take this problem which is a core problem within psychology and begin to approach it with the techniques of molecular biology. And I would like to describe to you essentially two sets of experiments. One, an implicit forms of learning to look at the nature of the switch and then an explicit forms of learning to again look at the nature of the switch to see if these two dramatically different forms of learning in fact share similar molecular components when it comes to switching short-term to long-term memory. So I'm gonna begin with a simple form of implicit learning using as a test animal the inverted bit of plazia and then I will show you also some data on flies. This is the marine snail of plazia. This is a giant snail that lives in the ocean. This is the head of this animal. This is the tail of the animal. Many people in the audience probably can't tell the difference and many don't even care what is the head of the tail. We have used this animal because it's extremely beautiful. It's also extraordinarily intelligent as you'll come to appreciate in a moment. But in addition, it shows a variety of behaviors that can be modified by learning and it has a simple nervous system. It has limited number of cells. Many of the cells are large and they're uniquely identifiable so you can recognize them in every animal of the species and give them specific names. This animal has a large number of behaviors but we focused on only one. One of the simplest behaviors the animal has a very simple withdrawal reflex. The animal has a respiratory organ called the gill which is normally covered by a sheet of skin called the mantle shelf that ends in the fleshy spout called the siphon. If you apply a tactile stimulus to the siphon you get a brisk withdrawal of both the gill and the siphon. That's illustrated here. The gill moves back into the mantle cavity. This is like the withdrawal of a hand or a hard object. As is true with your withdrawal reflexes these simple implicit forms of learning which you and me do not involve conscious participation can be modified by several different forms of learning and each of them shows both a short-term and a long-term memory depending upon the number of training trials. I'm gonna describe to you studies of a form of learning called sensitization which is a particularly elementary form of learning. Sensitization is a form of learning in which an animal learns about the properties of a noxious stimulus. So if you give the animal a noxious stimulus to the tail a shock to the tail it will recognize the stimulus and now it will learn to enhance its responses much more powerfully to a tactile stimulus than it would have prior to its receiving the shock. So it learns to avoid any stimulus much more powerfully than it would because it's not frightened and it prepares for withdrawal and escape. Now the memory for this noxious stimulus persists depending upon the number of training trials you've given to the tail. If you give a single shock to the tail you get a short-term memory that lasts for minutes. If you give four or five shocks to the tail you get a long-term memory that lasts one or two days. If you give an inhibitor protein synthesis you block the long-term memory if and only if you give the inhibitor during or shortly after the training. You in no way interfere with the short-term memory. If you give more than five trials if for example you give 15 trials you get a memory that lasts for over a week. If you give 15 trials a day for four days you get a memory that lasts for several weeks. So the duration of the memory is the function of the number of training trials that you give. And again in each of these cases you can show the long-term memory requires protein synthesis during and shortly after the training. Now because the animal is relatively simple we can work out in some detail the neural circuitry of the behavior. And this schematic diagram on the top shows you in very, very simplified form what the wiring diagram for the behavior is. Here is the gill and here is the siphon skin. The gill is innovated by six motor neurons that make direct connections to the muscle. They in turn receive information from the siphon skin by means of a population of sensory neurons that make direct connections onto the motor neurons and cause the motor neurons to fire. The sensory neurons also connect to inhibitory and excitatory interneurons that modulate the firing of the motor neuron. So essentially you have sensory neurons connecting directly to motor neurons and indirectly to two classes excitatory and inhibitory of interneurons. When the animal learns something when you give it a noxious stimulus to the tail you activate a group of facilitating neurons which are very much like the modulatory neurons in the brainstem of your my brain. The serotonergic, the dopaminergic, the cholinergic neurons. In the plysi of the major system is serotonergic. There is another system that uses a peptide as a transmitter and there's a third group of cells that uses a transmitter that's not yet been identified. These three groups of cells end on the sensory neurons including on their presynaptic terminal and there they act to strengthen the synaptic connections between the sensory neurons and their various target cells. If you give a single stimulus you produce a transient strengthening of synaptic connections that does not require new protein synthesis. If you give four or five stimuli you produce a long term strengthening of these synaptic connections that requires protein synthesis, new protein synthesis during and just after training. So we'll show you later on. This strengthening of synaptic connections occurs because more chemical transmitter is released from the presynaptic terminals after learning than is before. I will in a moment show you some of the molecular underpinnings of this mechanism of strengthening synaptic connections. There are interesting and important changes that occur at other points within the neural circuit but I'm only going to focus on this simple set of connections between the sensory neurons and motor neurons for two reasons. One is we find that this elementary connection between the sensory neuron and the motor neuron a representation of both the short term and the long term process. Since scientists are reductionists they want to understand things on the most elementary level. We thought we would use this as a beginning step for trying to understand somewhat more deeply how short and long term information is stored at the cellular level so we can see how information moves from short term storage to long term storage. Number one, the representation exists in the cellular level. But in addition we found that we could take these cells out of the animal and put them into dissociated cell culture. So together with Sam Shaker and Columbia we've been able to take a single sensory neuron in a single motor neuron and put them into a petri dish and grow them and they form perfectly good synaptic connections. You can now take a single serotonergic cell and plate it together with the sensory neuron and the motor neuron. If you stimulate the serotonergic cell once you get a transient strengthening of synaptic connections just like in the intact animal that lasts for minutes. If you stimulate it four or five times you get a strengthening that lasts over a day. As I'll show you in a moment the short term change does not require protein synthesis. The long term change does require protein synthesis. So here we've been able to reconstitute in a three neuron culture dish a component of the reflex that shows a representation of the short term and the long term change and really begin to explore with the tools of cell and molecular biology. In fact, you don't even need the serotonergic cell. You can do the experiments with just two cells in the culture because you can substitute a microelectrode that puffs out serotonin and just put serotonin onto the connections between the sensory neurons and the motor neurons. And this simply shows you what an experiment like that looks like. If you puff on serotonin on the connections between sensory neuron and the motor neuron this bar histogram shows you that you get a strengthening of synaptic connections. You get about an 80% increase in the strength of the synaptic connections. This lasts a period of about 15 to 20 minutes. And it is in no way affected by inhibitors of protein synthesis or RNA synthesis. Anisomycin, emetine, you don't need to know the names. Two inhibitors of protein synthesis have no effect on this at all. And two inhibitors of gene expression act in a mice and alpha manate have no effect. If you now look at the long-term process if you carry the culture for 24 hours you just get a depression of synaptic strength. If you take that culture and first give it five puffs of serotonin you get an enhancement of synaptic strength that lasts one or two days. But if you give inhibitors of protein synthesis you block this enhancement completely and if you give inhibitors of transcription that is of gene expression you block it partially. So here we can show on the cellular level that the requirement for protein synthesis is in fact not an emergent property. It is a property of the cells specifically involved in information storage and you can trap that requirement into two cells that you have in culture. In fact you can do more than just trap that requirement. You can show that that requirement has the exact time window, the consolidation phase that characterizes the whole behavior. Here I show you tongue-in-cheek, the time window for the goldfish experiment that I showed you earlier which involves the whole animal. And a similar time course is characteristic for the obliquia learning in the whole animal. And here I show you the time requirement in the connections between the sensory neuron and the motor neuron, two neurons in culture. If you give the inhibitor protein synthesis two or three hours after, excuse me, you've given the pups of serotonin, you have no effect on the long-term facilitation. In order to block the facilitation you have to give the inhibitor during or just after the training, the first hour or so. So this shows that in fact on the cellular level we have this time window and we can see what the specific genes are that are turned on and what the nature of the switch is that turns them on. And we're in a position to see whether we can detail the nature of the switch. This cartoon, as David Tuva would refer to it, is a very simple outline of the sensory neuron. This is the pre-synaptic sensory neuron, this is the motor neuron, this is the nucleus of the sensory neuron and this is the pre-synaptic terminal. If we stimulate the tail, we activate these facilitating neurons of which I've only drawn one ear, the serotonergic cell. And that activates this switch that I'm gonna return to in a moment which in turn engages two classes of genes that we've been able to delineate. One is a molecular machinery which turns on a specific protease which acts on one of the steps involved in the initiation process to keep the short-term memory going for a period of several hours. I'm being specifically vague about this step, not because we don't understand it, we understand it quite well, but because I don't think you need to focus on it right now. What I want you to focus on is the process that carries the memory not for a period of 10 to 15 hours, but carries it for a period of days. And that is that one of the things that switch for long-term memory does is to turn on a second class of genes which gives rise to the growth of new synaptic connections. And what carries the long-term memory, what is the signature of long-term memory, the characteristic feature, if you will, is the long-term memory involves an anatomical change, a growth of new synaptic connections. And I'd like to take a moment to just demonstrate this phenomenon for you. This is a home out, if you will, a fill with an electron-dense dye and a reconstruction drawn of it, of individual sensory neurons from animals that are naive and from animals that have undergone long-term sensitization training. These are experiments that Craig barely carried out at Columbia. He took two groups of animals, he produced long-term sensitization in one and nothing in the other, and then he took out the ganglia 24 hours later and injected individual sensory cells with a marker that allowed you to fill the whole sensory neuron. And this is a control cell that has not experienced long-term sensitization and you see the processes coming out from it and compare that to the animal that has undergone sensitization, long-term sensitization. You actually see a sprouting and outgrowth of new processes. Craig barely has been able to count the number of synaptic connections, both in control and long-term sensitized, and shown that with long-term sensitization in the intact animal, you get a doubling of synaptic connections. So a single sensory neuron normally has about 1200 synapses. After long-term sensitization, it has about 24 to 2,600 synaptic connections. And the motor neuron sends out processes to meet those incoming synaptic connections. Now this is not a casual observation because it implies that if you think back, and you remember David Chuball's lecture from yesterday, and there isn't a single person in this audience who heard that lecture, who's likely to forget that lecture, it is because there are structural changes I would predict occurring in your brain. So you walked out of this amphitheater yesterday with a somewhat different head than you walked into the amphitheater with because of the anatomical changes that have occurred in your brain. Now let me just modify that statement in some way. In this animal, we produce circumstances in which the changes are particularly dramatic. I doubt whether Chuball's lecture produced quite that large a change in your brain. Number two, some people worry. When they see that this kind of a growth occurs in the brain, they worry that their brain is gonna be filled up with anatomical detail, not to worry, don't worry. What has been able to show in the plesia that this anatomical change tracks the memory. Many memories do not last a lifetime of the organism. Your first love experience probably produced this kind of an anatomical change or more. That you will remember as long as you live. But David Chuball's lecture, let's say 10 or 15 years, you will retract those connections with time. In addition, there are learning processes that involve not the growth of connections, but retractions of connections that begin with. This animal shows a learning process called a situation in which it learns to ignore stimuli in its environment. There the synaptic connections go from 1200 to 800. So not every learning process needs to a growth, but all long-term memory leads to a structural change. Now clearly we would like to know, how is this switched on? And my colleagues and I have spent a fair amount of time trying to analyze the detailed steps whereby the structural change is switched on. And I'd like now to sort of take you through the experiments and our current thinking about how it works. If you stimulate the tail and activate the serotonin cells, these serotonergic cells release serotonin and serotonin acts on a serotonin receptor that activates an enzyme called the serotonin-sensitive adenyl cyclase. The enzyme adenyl cyclase converts a common molecule of ACP into a messenger, an intracellular messenger called cyclic AMP. And cyclic AMP has the property that is capable of activating another enzyme called the cyclic AMP-dependent protein kinase. Protein kinases are proteins that put phosphate groups on other proteins. That's a mechanism whereby you can take a protein that is asleep and wake it up. It's a way of turning proteins on a roll. And it activates several sets of proteins in the cytoplasm of the cell. For example, it acts on potassium channels that increase calcium influx into the presynaptic terminal to enhance transmitter release from these terminals. And it also acts on the machinery for transmitter release itself. When you give a single pulse of serotonin, you increase cyclic AMP and you increase the activity of the cyclic AMP dependent protein kinase in the cytoplasm. But there is no cyclic AMP and there's no protein kinase in the nucleus. Roger Tien, a very gifted investigator has developed techniques that allows you to look within the cell where the cyclic AMP dependent protein kinase is located at any given moment in time. And we have collaborated with them and we have found that when you give a single pulse of serotonin and you activate the protein kinase, you activate it only for a very, very short period of time. And in that period, the protein kinase is only in the cytoplasm. But if you give four or five repeated pulses, the protein kinase moves into the nucleus. It translocates into the nucleus and there it activates genes, including the genes that give rise to growth as well as the genes that I told you about before that make the kinase persistently active. So even in the absence of cyclic AMP, it will continue to fast correlate progress. This was the step that I told you about before. Now, what does it do in the nucleus? In order to understand what it does in the nucleus, I have to explain to you in one or two words how genes function. Genes encode proteins and proteins make you what you are. And the genes that make proteins have two parts to them, two segments, two DNA segments. The major part of the gene has a coding region which encodes the sequence of nucleotides which codes for the protein. But in addition, there is to the left, molecular biologists call it upstream of the coding region, a piece of DNA that serves as a control region that determines whether a given gene is gonna be transcribed, whether it's gonna be read, whether it's gonna be eligible for action. This control region is often called the promoter or enhancer region, works by having specific proteins bind to it. So the switches are the proteins that bind to these DNA control regions. There is a specific protein that responds to cyclic AMP and that protein binds to a specific component of the control region. That component of the control region that recognizes the protein that recognizes cyclic AMP is called the cyclic AMP recognition element and the protein itself is called the cyclic AMP recognition element binding protein. There's nothing to molecular biology except a few English words. What the kinase, the cyclic AMP dependent protein kinase does is translocate to the nucleus. It phosphorylates this protein, it activates it, so it can act on, it can be recognized by and act on a control region that controls genes that are responsive to cyclic AMP. Now there are many genes that are not responsive to cyclic AMP. They do not have a cyclic AMP recognition element in the upstream region or it's so completely covered up that it can never have access to protein. Those genes will never be activated by this protein. Genes that have the recognition element will only be activated if and only if this protein, the cyclic AMP recognition element binding protein CREB can bind to it. This would predict that if in fact you can block this protein from binding to the recognition element, you could selectively block the long-term process. And my colleagues and I have carried out a number of different experiments that show if you block the binding of CREB protein to its normal enhance elements in the sensory neurons, you can block the long-term facilitation. One experiment we've done is to synthesize an artificial piece of DNA that encodes the cyclic AMP recognition element. So we can inject that into the nucleus in excess, in million copies. Now this protein doesn't know what to do whether to bind to the native gene or to these extra delicious alien enticing seductive pieces of DNA that we've injected in the nucleus. That protein ain't no fool. It knows where to go. It goes to the alien element that you've injected and it blocks selectively the long-term process. This is simply shows you the long-term facilitation you get with repeated pulsus serotonin and the number of synaptic terminals that increase in culture as a result of that. So this increase in number of synapses accounts for this increase in synaptic strength. If you inject this piece of DNA, you selectively block the increase in synaptic strength because you block the increase in number of varicose. And we can show with a number of controls that injecting control regions for other genes has no effect and the short-term memory is a no-way alter. I want to return to this cartoon for two reasons. The first thing I want to show you that I've shown you really the bare-bones scheme as we understand it. We have a much more detailed understanding. We know much more about additional steps involved. We know exactly the steps involved in this molecular machinery that gives you the intermediate step of memory. We've identified a very interesting gene that serves as a bridge between CREB and the structural change, a gene called CBP, and for the transcriptional jocks in the audience, let me say this is an immediate early gene that has all the features of force in June. And we've shown that just as you can block the long-term facilitation by blocking these genes, you can also block the growth by selectively blocking this one gene. What is fascinating is that this organism has 10,000 genes. You can block either this gene or that gene that is one gene at a time and selectively block the long-term memory process. The second reason I want to show you this slide is I want to emphasize aspects of the generality of this model. One of David Chuball's colleagues, Marge Livingston, in an early incarnation, actually worked with fruit flies. This is before she got lost in the visual system. And she carried out some very beautiful experiments in fruit flies in which she showed, earlier, Benz's group and some others had shown, that fruit flies, Drosophila can learn, you can create single gene mutations, one gene out of 10,000 to fly. Knock it out, the fly is dumb, can't learn. You have flies like dunce and rutabagger, various vegetables that have memory defects. Two of these have been analyzed in some detail and each of them involves one or another step in the cyclic game PKS game. What is really fascinating is in the last several months, people have shown that these fruit flies not only have learning and short-term memory, they have long-term memory, which was not appreciated before. That long-term memory requires new protein synthesis, just like a plesidae learning, just like all kinds of implicit forms of learning. And they've identified a crab protein in flies, which is no surprise, but they've shown that if they block the crab protein in flies, they selectively block in the intact fly, long-term memory without short-term memory. And this is experiments that were carried out by Jerry Yin and Tintelli and Chip Quinn, and this shows that with a certain training paradigm repeated space training, you get this is an olfactory discrimination learning, you get a nice long-term memory that lasts a week, this long-term memory is blocked by inhibitors of protein synthesis. You can produce a manipulation that will selectively block the crab gene. Technically what this involves is that there is a gene that prevents crab from acting, it's called CREM, and you can put that under the control of an element that is sensitive to heat. So if you give the animal a pulse of heat, just heat it up for a moment, you can turn on this gene selectively. And this simply shows the block of this long-term process. This is just giving the heat shock to the animal that doesn't have the construct that shuts off crab, and this is giving heat shock to flies that have the construct that selectively shuts off crab. So you selectively block this short-term process, the long-term process, you have no effect on the short-term process. So grant that this is really only two kinds of learning, implicit kinds of learning in two kinds of flies, but it's encouraging to think that CREM is an important component of a switch that has some generality. Now, to what degree does it apply to explicit forms of learning? To what degree does it apply to your brain and mind? This is an MRI of the human brain, and this is simply to show you what the mesial part of the brain looks like, and particularly the hippocampus. This is from Larry Squire and David Amaral. The hippocampus, which is the size of the thumb, sits in a little shelf called the sebiculum, and it was really Tony DeMarsius' group that first pointed out that the sebiculum is absolutely critical to provide the input to the hippocampus, and this sebiculum is particularly susceptible to damage during Alzheimer's disease, as is the hippocampus because the sebiculum, the input is damaged. Small structure, the size of a thumb, if you damage it even in one segment, you have severe loss of explicit forms of memory. What is it about the hippocampus that makes it such an important site for information storage? Well, the honest answer is we don't know, but we're beginning to get some very good clues. This is a cross-section of the hippocampus. The input from the entorhinal area comes in and synapses in a group of cells in the hippocampus, and these cells synapses in another group of cells, which synapses in a third group of cells. So there are three sets of synaptic connections within the hippocampus, and they all have fancy names and they're not of any importance to you. What is interesting is that Bliss and Lomo, two investigators working in the 1970s, first discovered that each of these synaptic connections has an interesting property. A brief period of activity in any one of those synaptic connections, and you get a long-term change in synaptic strength. That is called long-term potentiation, and the number of investigators, including Suzuma Tanagawa and my colleagues and I have shown that if you knock out genes that interfere with long-term potentiation, you interfere with memory storage. So we've been interested in asking, does long-term potentiation have a short-term and a long-term phase? Is there a cellular representation of these two components here as well? If so, does the long-term phase require transcription? Does it invite gene activation? And if so, does that gene activation require cyclic AMP encrypt? The first thing my colleagues and I found in the hippocampus, as in the plesia, practice makes perfect. If you give a single train to any one of these pathways, you produce a strengthening of synaptic connections. This is a time course that graphs, this is a graph that plots the time course of the enhancement of synaptic strength following a single train. And you see you get an enhancement of synaptic strength and we followed it, it lasts about a couple of hours. It's a short-term process, and it's in no way affected by inhibitors of protein synthesis. But if you give three, four, or five trains, you get a process that lasts for many, many hours, and this new component that is added, the new long-term component, requires new protein synthesis. Not only does it require new protein synthesis, but it requires gene expression because you can block it with inhibitors of transcription. Here, we give repeated trains and you get this nice long-term process. If you give an inhibitor of transcription, you get the short-term process, but you block the long-term process. But the nicest thing is that there is, on the cellular level here, the explicit forms of learning, as there is, in implicit form of learning, a cellular representation of the consolidation phase. If you give the inhibitor of transcription one or two hours after LTP is started, you don't block anything. You continue to have it. Transcription, the activation of genes, is important during and right after. It's only if the ectenomycin is present at the onset of the train and for the first hour or so, you block the long-term memory. This, of course, caused us to ask, is this late phase mediated by cyclic AMP? We could show, we could simulate it perfectly. And it raised the question, are there genes that are induced by cyclic AMP that might contribute to this late phase? And my colleagues and I searched for genes that are induced by LTP and by cyclic AMP. And we found a number of genes, which we've called the bad genes, brain activity-dependent genes. And this was actually, these genes were found by Joe Kwan in my laboratory, whose advisor is sitting in the audience and he has the right to be proud of them. One of the genes that he identified is particularly interesting and I wanna focus on it. This gene is TPA, tissue plasmidogen activator. Now, many of you may know of that gene because this is one of the wonder drugs that are given to patients who suffer the acute heart attacks because it clears the coronary arteries. It's a protease that clears away other proteins, clears away the clot. Now, the interesting thing about tissue plasmidogen activator, it is secreted by neurons doing the process of growth. So it clears away the debris and the extracellular spacial process can grow out. And this TPA gene has a region upstream, which is a CRE and it binds a CRE protein. So this is the coding region that encodes this protein below here and here is the regulatory region. And for those of you who are interested, you see there's not just one control region, there are control regions for different binding proteins so other kinds of events could also activate or shadow off these genes. But we were interested in seeing how this gene is activated by psychic gain. So we were very fortunate at this particular point because Richard Mulligan at MIT generated a mouse that had every gene intact, every one mouse now has 100,000 genes comparable to the number of genes that you and I have knocked out selectively this one gene for TPA. Also knocked out in another strain of mice a closely related gene, which encodes for another protease, UPA. We took the both strains of mice, the UPA mouse that has nothing to do with LTP had no disturbance whatsoever in LTP. But if we looked at the TPA mouse, we found the following. When we gave a single trial to an animal that was lacking TPA, we found there was no effect whatsoever. LTP, the short term LTP was completely normal. These are superimposed graphs of a wild type mouse in the mutant mice and you can tell the difference. But if you give four repeated strains, you can get the long term process in the mutant. This is, I'm sorry, in the wild type in the normal mouse, this is what happens in the TPA mouse. There's a severe depression of the late phase. Not a complete abolition, but a severe depression showing that knocking out this single gene has a dramatic effect on the late phase. So these experiments begin to suggest that in explicit forms of learning, in the hippocampus, a KREB-like switch is involved just like in implicit forms of learning in Drosophila and Eplisida. This slide is a cartoon that summarizes our understanding on the molecular level of how things seem to work in LTP. This is the pulse synaptic spine of a synapse and this is the presynaptic terminal and in the same neuron, this is another pulse synaptic spot. When you activate the synapse repeatedly, brrr, you activate a specific class of receptors to this transmitter. This uses glutamate as its transmitter and you activate a specific receptor called the NMDA receptor. It's really not important. What is important about this receptor is that it allows calcium to go into the neuron. Calcium is like a cyclic amperes, it's a very important signaling molecule. It activates two processes. It activates a set of short-term processes that are completely different that are involved in Drosophila and Eplisida. They also involve protein kinases but they involve completely different protein kinases. Protein kinase C, calcium, calmodulin kinases, they're unimportant. The important thing is that the short-term process works by a completely different mechanism. But calcium also acts on the adenyl cyclase in these neurons, the enzyme that synthesizes cyclic amperes and we've shown that it synthesizes cyclic amperes that activates genes that give rise to the long-term process. So even though the short-term process is completely different in this example explicit learning compared to the implicit learning I've showed you before, these preliminary data suggests that in both cases, the long-term process that is switched on seems to share a common component of the switch. Now clearly to get rigorous evidence for this, you would like to do several things. One is you'd like to begin to explore behaviorally what happens in these TPA knockout mice and we're beginning to do that. You would make the prediction that they lack long-term memory but have perfectly good short-term memory. Specifically for explicit forms of learning but you'd also like to knock out the Kreb gene. Now it turns out that just recently somebody has knocked out the Kreb gene and Alcina Silva and his colleagues at Cold Spring Harbor have looked at what happens in the mouse using this kind of a paradigm when you have a selective knockout of the Kreb gene. They have found that it interferes with LTP and particularly in the late phase and they have done behavioral experiments. Uthiga, Uthigana in Alcina Silva's laboratory has done behavioral experiments to show this is a spatial task, an explicit task which animals learning its orientation in space that the animal learns this task perfectly well that it has almost perfect, not quite almost perfect short-term memory but if you look at long-term memory after one hour looking at one hour and at 24 hours you see a dramatic difference between the wild-type mice and the mutant mice. So knocking out Kreb in a mouse shows an interference in an explicit learning task which is not evident in the short-term process, not evident in the learning process but evident in the long-term memory. So this allows me to sort of summarize the several points I've tried to make here and to try to extend it in a sort of speculative way to some broader perhaps more philosophical issues. Let me summarize by saying that molecular biology is beginning to bring together cognitive neuroscience and signaling aspects of neurobiology and it is possible to begin to think in very primitive terms of perhaps developing a molecular biology of cognition and that particularly using transgenic mice in which by knocking out a gene you can look at the animal's physiology while at the same time being able to have an intact animal whose behavior you can explore you can relate single genes to synaptic events in the brain to whole animal behavior. These kinds of studies put together with studies in advertebrates indicate that even though learning and memory are extremely complicated processes and fall into two major classes implicit and explicit and have completely different rules both share similar molecular mechanisms for memory storage. In both explicit and implicit learning you have short and long-term memory. In both of them you have a cellular representation of both the short-term and the long-term form on the cellular level. The short-term form involves activation of protein kinases. The long-term form involves activations of genes and structural changes and the switch from the short-term to the long-term form. The characteristic feature of the consolidation phase is the activation of genes by means of a KREB-like protein. Now, what does this mean for the person on the street? What does this mean for you and me? I'd like to leave you with some general ideas. This is you and me when we really come to honest grips with ourselves. This homunculus is a representation of your body form on the surface of your brain. This is what Tony Odomasio, David Tuval, and Patricia Church didn't look like on the surface of the brain. Obviously, Patricia Church is much more handsome than that. Each of us has a representation which is real but distorted. Those parts of the body that are particularly important have a larger representation. That is, your body's surface can be mapped in the surfaces you bring. In this amount of sensory cortex, there is a representation of every part of your body. And hands and face, which are important as tactile organs, have a particularly large representation. Now, Mersenne at the University of California, San Francisco has been interested in the nature of that representation. And he examined the representation of a bunch of monkeys. And he found that different monkeys, when he examined them, had different kinds of representation for the hand. They differed somewhat in the shape and the size of the representation. Now, it was difficult to know from those experiments whether this is due to genes or difference in experience. They didn't have a Minnesota experiment in which they had monkeys, identical twin monkeys, that had been separated at birth so they could examine them 20 years later. They still need to do that. I told you, California's way behind Minnesota. So he did the next best thing. He took a monkey, exposed its cortex, and mapped its representation for the fingers. And here are the five fingers of the hand. And this is the representation for each of these fingers. This is the thumb and this is the pinky. I'm sorry, thumb and pinky. He then had the animal, not play a piano, but the next best thing for a monkey. He had him press a ball for food many days, and the only source of food the monkey could get was by pressing a lever with three fingers of the hand. And then he mapped, after several weeks of having the animal press a ball only with these three fingertips, remapped the surface representation. And what he had found, and this is not being confirmed by a number of other people, is that the representation of those three fingers had expanded at the expense of other areas in the brain. This is really another way of saying what I told you before that when you learn something, in this case an implicit task, it produces structural changes, and you can really map that in the surface of the brain. And although there is no evidence for this, my guess is, and one really needs to test that, that this is due to a structural, this structural change is due to alterations in gene expression. So the argument that every major event that you remember is likely to leave some anatomical traits, and that anatomical trace is likely to be to alteration in gene expression, gives you a view of mental processes that I would like you to take home with you if you will. It allows you to look at mental illness in a somewhat different term, and this is not unique with me, this is an argument really many of us in the neurobiology community feel. It makes you realize that genes are important not only insofar as they're defective, but even normal genes are important for everyday behavior because they can be switched on and off. What I'm gonna show you is that even though there are dramatic differences between psychotic illnesses, which have a major genetic component that is heritable, neurotic illnesses, which to some degree are not inheritable at all, they can maybe completely acquired, also involve genes, although not in a heritable way. If you take a mental illness like schizophrenia and depression, which we know from good familial studies, involves alterations in genes, mutations, not a single gene, but probably a number of little genes acting together to produce a chronic depression or a schizophrenic syndrome. That leads to a mutation in specific sets of genes that produces an abnormal RNA or no RNA that produces an abnormal protein or no protein that leads to the disease. And this abnormal gene, this mutated gene will be present in every cell of the body, including the egg and the sperm, and therefore will be passed on from generation to generation. But the data on learning indicates that things like anxiety states, various kinds of neurotic illnesses, the garden variety of neurotic illnesses that you and I suffer from, which can be induced by environmental stimuli, might involve something like a learning process, activating genes that are there and normal, but that may be shut off. And now activating them gives rise to a protein that produces a chronic anxiety state. Clearly, this is an outline of a possibility, this is not an established, but the studies that learning make you think that in fact, learning events can also lead to abnormal processes, and these abnormal processes are likely to be represented in structural changes that are produced by the switching on and off of genes. What is the consequence of this? One consequence of this is that insofar as neurotic illnesses involve alterations in gene expression that lead to structural changes, it is possible that as Hannah DeMascio gets more and more gifted in her imaging resolution, she may be in a position to begin to detect even these kinds of anatomical changes that accompany learning processes, the learning of abnormal experiences that accompany neurotic illnesses, and the improvement of neurotic illnesses by counseling and psychotherapy. So I would wager to guess that 50 years from now we may have resolution enough to see, number one, the nature of neurotic illnesses by looking at PET scans or MRI, what's the 20th generation version of those instruments, be able to detect anatomical changes that are associated with these neurotic illnesses and follow the outcome of drug treatment and psychotherapy to see whether they're really bringing out the anatomical changes that reverse the process. Steve Rasmussen at Brown has recently shown me some data that with obsessive compulsive neurosis, you in fact can detect a specific anatomical change in the frontal cortex, an area that Tony was talking about being very important for the integration of emotional and cognitive experience. With specific serotonin uptake blockers, that abnormality is reversed and psychotherapy also reverses that abnormality and it also produces anatomical changes. Now this is the initial study, one needs lots of replications before one can believe that. I only hold it out to you as a way of thinking about the possibility that neurobiology holds for understanding these various kinds of illnesses because I would guess that a president at this college 20 years from now will be able to get an idea of whether or not his curriculum is really having its effect on the student body that he thinks, thinks it's having and also we can see whether students have effect on the faculty. Thank you very much. Or was it too abstract? No, no, not at all. It was a very insidious one. It was a sort of, you know... I'm going to stay with the mics. Okay, the clinical situations. No, it's definitely... I think they made over to behavior modification. They made over to behavior modification. Yeah, yeah, okay. If you play a week, I would... Yeah, what else? Speaking of this... I wonder if I'll text the other then I can. I feel so privileged. Speaking for eight years or so. Yeah, I was thinking actually, but I don't think that he's speaking I think we'll go ahead and get started with the questions. Dr. Churchill will be joining us shortly. We'll start over on this other end here with Dr. Georgiopoulos. Well, I just want to start the comment if you want with the statement and the question, of course. The statement is that that was the most extraordinary lecture that I have held for a while. And it's not for the accumulation of facts. These we can all get easily. But for the thread that over so many years has run through Eric Handel's work in pursuing what is somehow it can be called in my hands and it becomes called in his hands which is a continuum in bringing together molecular events and behavior across species and across human illness. I personally I found the focusing down if you want or narrowing down the gap on the Kreb gene is extremely reassuring and quite disconcerting. I reassuring because it's an exciting because it opens the way for the union of molecular biology, behavior and cognitive science as Eric Handel mentioned. I don't know of any other single most important question in that route. I find it disconcerting and that's what I would like Eric to comment on because it seems to be a bottleneck. It seems to be a narrow gap through which a lot of aspects of learning and memory, especially what is retained not what is being learned from moment to moment seems to all of these processes seem to have to go through. And what I find somehow disconcerting is the possibility if we deal with the narrow gene or the family of genes that are crucial for this role and which with a simple knockout mice model you can produce rather profound effects both in cellular function and behavior. Where do we stand in terms of environmental hazards that could potentially interfere with such genes? Toxins, food additives I don't want to pinpoint anything in particular but we have been exposing ourselves manufacturing most of such materials and somehow we always thought the brain is very redundant. There are so many different ways you can do this and that that why we are known we lose so many thousands brain cells a day and if you lose a little more one will still do it survive and function the bottleneck of a narrow family of genes that seem to be extraordinarily important for such functions as learning and memory I find where we're exposing ourselves to a vulnerability that we may now that we understand that or start to understand we may also start to be more careful and somehow look after ourselves protecting ourselves I don't know what Eric might want to comment on that I think Pastor's comments are very well taken I would really only elaborate on two points that I didn't make in my talk one is that in the mammalian knockout for example of KREB people didn't appreciate it when they first did the knockout but there are actually two different isoforms of the KREB gene that are transcribed from the same gene and the knockout only succeeded in eliminating one of those so the animal still has another isoform this is not present in the brain but it's present to other cells in the body number one so there is for many genes a redundancy so that you could affect you know, regulate one but not others there's also another point which I really fail to mention which is really essential to mention and actually David Jubil wrote about this at one point a number of people have commented on it there are really a number of different problems with learning and memory one is the synaptic mechanisms that are likely to be involved in it and that in some ways is the easy part of the job although we are by no means understand it completely but you also now have to take that mechanism and put it into the kind of complex circuitry that you heard about yesterday I mean there is not one synaptic relay that changes in a complex learning task but many so there is also opportunity for redundancy in the wiring diagram and also as you go along the wiring diagram you may get changes in synaptic strength in some components and not in others so there is the spatial distribution of that learning principle if you will that needs to be taken into consideration what I think is reassuring is that despite the fact that we still have to do the hard work of figuring out how does the information get into the hippocampus how does it get out of the hippocampus etc it the molecular analysis of what happens at each site may be simplified if it proves that there are only a limited number of mechanisms and I certainly wouldn't want you to go away thinking that this is the only way long-term memory can be stored I don't for a moment think that I think this is a common mechanism I do not for a moment think it is universal Dr. Sacks I found your presentation very fascinating indeed because of the sense of a universal mechanism intermediating between short long-term memory throughout the animal kingdom I am wearing a gastropod tie in your honour and I think there is a certain danger for clinicians and for that matter sort of vertebrate specialists to look down our animals I certainly spend a lot of my time in zoos and aquaria and even botanical gardens because I have a very strong sense of the unity of life and I think DNA studies are showing this increasingly that we are not that far from a CQC or for that matter from a rutabaker something which I found enormously interesting in my student days was Jay Z. Young's work on learning in kephalopods and his model of the brain was of course built on that by the way I did not entirely like your comment about rutabakers and I wondered whether in the sensitive plant in Dionea which seems to show some ability to learn an ability to not respond to raindrops and other non-significant stimuli where the one has anything like this but most of my thoughts were of a clinical sort three in particular I was fascinated when you made a connection with Mersenich's work on the plasticity of body image and having sort of broken my arms and legs and been variously operated on and casted I've had quite a lot of experiences of abnormalities of body image in particular it seems to me that if a limb is immobilized or prevented from action for more than two or three hours one gets a massive change in body image I had that myself when I had otherwise identical leg injuries with the left leg I was really prevented from movement for about two weeks and during that time I had a severe disturbance of body image I really lost the memory of the leg whereas when this was done with the other leg and I was permitted to use it in a walking cast two hours after operation there was no hiatus in memory it seems to me that within an hour or two of lack of input from the body one starts to get an oblivion and a change in body image and this is one thought secondly I noted you mentioned emetine in one of your cartoons I don't know whether this is the same as apomorphine but anyhow I have I saw an interesting a Parkinsonian friend and patient of mine as he was freezing up himself an injection of apomorphine subcutaneous injection this can work wonderfully and really within about 60 seconds he straightened up and smiled and what he said was I have forgotten how to be Parkinsonian although he said that the terrible memory of how to be Parkinsonian would come back as soon as the apomorphine wore off and I again wonder about this sort of neural knowledge or neural memory in your terms one final thing which came up when there was some talk about environmental poisoning as many of you know you better watch eating muscles sometimes if you eat muscles when you wake up the next morning you don't recognize your wife or anyone you've blown your hippocampus the muscles are full of acidic acid but anyhow these were some thoughts as you were talking thank you it's hard for me to elaborate that a body image is a fascinating thing and that would be very worthwhile studying because that would provide a very nice time period in which you might be able to look for changes in brain structure as a result of that and certainly would encourage as physicians do the mobilizations of patients in a hurry in order to prevent those drastic change the current but you may also remember that the more dramatic the de-affrentation is in the PONS experiment in which monkeys were de-afferent it was extraordinary how dramatic the central representation had changed also with phantom limb there is now evidence for the fact that the you have a other parts of the body if you stimulate them can give you the sensation of feeling the particular fingers for example missing on the hand and I think the third example that comes to mind isn't one the best documented one which is syndactically the children that are born with the fingers of the hand fused those children have a very tiny hand if you look at the surface representation it is very small and there is no order to the stubs of the fingers when surgeons separate the fingers the hand area the representation grows quite dramatically in the surface of the cortex and now it assumes the same order that it does in the normal hand so here people have been able to follow in a living subject the change in representation of the body image if you will as a result of surgical improvement of syndactically Dr. Churchill I just wanted to ask you briefly if you might comment in which your approach and that of Jerry Adelman are similar in a couple of respects in which they differ he has suggested that by and large the learning consists in selection as opposed to instruction and that the main function is a pruning back as opposed to an activity dependent reaching out and yet I notice for example that you do in your slide the motor neurons you indicated that the processes grew towards a certain direction so assuming that that is a useful distinction between sort of instructional growth and learning and selectional growth and learning can you just say a bit about where you stand relative to his approach yeah first of all there is unequivocal evidence about the fact that you get both growth and subtraction of connections and the retraction takes several forms one is if you have a memory that lasts for several weeks as tested behaviorally and you look at the sensory neurons as a function of time that say every five days you see there is a progressive retraction of the number of synaptic connections that you can count and similarly with things like long term obituation you can see a retraction and the big mistake between selection and instruction is do you grow connections between neurons that were previously never connected or is the growth an amplification of connections of neurons that are already hooked up together and we can't answer that question the basis of what we've already done although in principle it's answerable what we have done is we've looked at cells that we know are connected and we've strengthened and then we've correlated that with an anatomical change that parallels it we need to look at two cells that are not interconnected to see whether there is a result of this great growth the sensory neuron now forms connections with cells that it normally does not form connections with and that's perfectly possible giving the size of this growth and that would distinguish between instruction or selection that is the selection would say the learning process instructs a new set of connections I think the one issue that actually addresses Jerry Edelman more it's still unclear although I must say his thinking has been helpful in this regard is to what degree learning phenomenon are important in development now it isn't true in the primary visual cortex but there is evidence in a number of systems where learning like processes are important for synapse specification that is Spurry originally had the idea that when the brain develops there are three sets of cues that are chemically guiding the outgrowth of the axon defining the appropriate target and the honing on the precise specific postsynaptic cell there are now a number of studies that suggest that the first two are chemically coded and require specific recognition molecules but the actual fine tuning of the synaptic connection the honing in a specific target involves an activity dependent process akin to learning now as David pointed out that need not be an environmentally programmed one then as Carla Schatz has shown very beautifully that the closed eye which has no visual experience whatsoever nonetheless has in it brittle ganged in cells that are spontaneously active now when you follow those into the lateral geniculate they use a mechanism for fine tuning the synaptic connections which is very similar to that which is used by learning so it's a learning rule but it is not an experiential learning phenomenon that is doing it but it is a continuous activity go to one more very brief question from Dr. DiMazio this is not really part question part comment yesterday when at the beginning of my talk I mentioned the excitement that we are now experiencing this perspective that is opening of linking up the molecular level of structure all the way into what is traditionally known as psychology I was actually thinking about precisely the kinds of things that Eric Kendall showed very beautifully this morning and when you look at the the way in which he is approaching distinctions such as explicit and implicit memory in terms of their shared neural base and the comments he made about short term and long term molecular level processes you realize how exciting this is and the perspectives this opens now I wanted to make one question for you you said at a certain point when you talked about especially long term memory you mentioned that in fact you did not have to conceptualize long term memory as an emergent property of the system and I totally agree with you however there comes a point at which there are characteristics of the system and operations of the system that could not be predicted at all from the molecular level one example is that at the molecular level will operate for long term memory say in relation to unique faces will also operate for non-unique faces or for a manipulable object or for pitches in a melody and the kinds of distinctions that we see for instance in some of the studies that we're making that we're doing rely on levels that could not have been predicted from the molecular so in a way in fact parts of the system and parts of the function that emerge at levels of high complexity and that need to explain that those levels although their base their grounding is still in the molecular and cellular structure you're absolutely right and I over simplified and I didn't mean that because I tried to point out here that behavioral memory is a systems property except in the very simplest kinds of reflexes what I meant is that you it wasn't absolutely obvious that we needed to have a representation of short term and long term synaptic transmission that parallels the memory in any way whatsoever or that the consolidation phase have a cellular representation you could have made it more complicated even though you've got this elementary component that doesn't mean if you put this in series in parallel you're not going to get some higher order phenomena and that probably explains much of the fascination of memory we're here at the very beginning and some things can be reduced others will require understanding of the whole system I have a few questions from the audience I'm going to group them so that you get some feel for the kinds of questions that folks have asked the first one I think I might even be able to answer it says in view of interruptability of short to long term memory conversion might sleeping in class prove useful as a tool to avoid further stimulation after learning a body of facts well the problem is knowing when to sleep and I've got a lot of direct evidence that it's not a very productive approach another related question is if memory is protein production does the brain gain weight in the process and what happens to the weight of the brain over a period of learning i.e. an entire college career I think people have actually done that study with animals and another related question has to do with how does the structure of the typical school day really facilitate the kind of consolidation periods that you spoke about one of the things about experiments in memory is that they're quite artificially constrained that is the best way to get an animal to learn something is to put an environment in which the only thing that is significant of the stimuli which it associates or the learning event that occurs we're embedded in a world in which there are lots of stimuli and the learning event is sort of the signal that emerges above the Norway's there haven't been that many studies of learning under those circumstances to see how one optimizes the learning situation I said it's a joke taking breaks in between one thing you know and that is again I have to go back to your mother I get the more I listen to my mother and that is there are studies in students which show exactly what you see in snails there's a difference between what is called mass training versus space training if you cram the night before the exam you will remember much of what you've learned quite well the next morning you will not remember it very well a week or two later but if you space the training that is do your homework every night it is surprising how much more enduring the memory becomes again there are exceptions to this but by and large space training produces much more effective long-term memory my friends in Drosophila neurobiology could not after 15 years of work demonstrate any long-term memory that lasted days and the reason they couldn't do it is because they were always using mass training putting the trials one after another Tim Tully's great contribution that led to that slide which I showed you came from the realization that space training might work and flies as well and got dramatic results so for example if you take 40 trials and give them in groups of 10 and separate them by two hours for a day you'll get much better learning that if you just take the 40 trials and run them right on one right after another so I think that's one thing that has emerged as a general rule another set of questions one from a hopeful questioner was is there a pill that one can take to produce the precursors that are necessary to facilitate long-term memory and a series of related questions that wanted to know whether or not there might also be these protein related deficits that might be predictive of problems in Alzheimer's or problems in other related illnesses yeah I mean clearly there are anatomical and molecular defects in Alzheimer's disease I'm not sure they're necessarily related to what I talked about these are degenerative diseases of the nervous system in which one loses some of the basic modulatory circuits it's a little bit like getting rid of the serotonin connections in the plizia for example you're losing a major cholinergic input you're losing nerve cells that are innovated by this cholinergic input so that's a more massive degenerative disease than I've been thinking about in my talk I think the kinds of mechanisms that I've been talking about if they have any immediate medical significance at all it is more likely to the kinds of things that Anders Bjorklund has been looking at which is memory loss in normal aging rodents in which he's been able to show that these animals some of them undergo memory loss with age and that can be restored by restoring certain of these modulatory factors and growth factors and we have inspired by his work just begun to take a look at what happens to LTP in aging mice and we find that it is this late phase that is selectively affected and we've used one of the agents that he has used NT45 and shown that it will reverse this late phase. Now this is a long these are very early experiments but the point I want to make is it is possible that over the next 10 or 15 years one might be able to develop a rational pharmacology of memory related of age related memory loss I think that is likely to be a simpler problem than the really more dramatic structural changes that occur with the Alzheimer's disease. I think that's that's a very difficult problem and I don't at the moment see a near solution to that. Thank you very much. At this point I'll remind you that we reconvene at 130 and ask you again to thank Dr. Kandel for a great thank you.