1999 LECTURE SERIES

Learning Retunes the Brain: The Neural Representation of Experience

Dr. Norman Weinberger
Center for the Neurobiology of Learning and Memory/Department of Psychology
University of California, Irvine
April 6, 1999

Our lives consist of experiences. Each of us is what we see and hear and the sights and the sounds located outside of our bodies. In the real world in which we live, how do we personally get the outside world into our minds? Of course, we can't get the actual outside world into our minds, it wouldn't fit. Instead, we construct a representation, a facsimile of the real world inside of our heads. When we think of all the many mysteries of the human mind, none is greater than that of how the outside world is recreated within ourselves. In everyday life, we don't think about this fundamental problem of brain function, because our experiences are so immediate, so vivid, and without effort that we just take them for granted. But this really is an enormously complicated problem. Think about it. When we see something, it doesn't go to a video recorder behind our eyes. When we hear something, it doesn't go to a tape recorder between our ears. Even if there were video and tape recorders inside of our heads, who would look at the tapes? Instead, there is a machine inside of our heads - the most wondrous and complex thing that we know of in the universe - the human brain.

The brain is where we see and hear all of our experiences. It is really amazing, almost unbelievable, that our brain somehow constructs a reality of the outside world inside of our heads. But the brain does something even more incredible. It actually retains some of these experiences for minutes, days, weeks or even lifetimes. The brain creates memories. So the brain really has two enormous problems. First, it has to provide a reasonably accurate picture of the outside world, inside the brain. Second, it has to store some aspects of these experiences more or less indefinitely. To make very clear the difference between these two brain tasks, think about your most recent experience. That's the experience of what I just said, which was "think about your most recent experience." The experience of first hearing that sentence is now in the past. It's a memory. A very recent memory, but a memory.

Our psychological present is very brief. The present is fleeting. The rest is all memory. Memories are the only experiences that we can hang onto. Think about what you had for breakfast this morning. You're recalling a longer-term memory than the memories of anything you've experienced in the past few minutes. Think about a teacher that you had in grade school. I remember a teacher in middle school named Bulldog Nelson. Children are cruel, but she was a wonderful English teacher. Our psychological lives consist almost entirely of memories, and our current experiences are called "percepts." Events in the outside world enter into our body, into our brains, and form percepts. A small fraction of those percepts become memories. So we do not remember everything we experienced, nor would we want to. As I mentioned, the brain has two fundamental and essential tasks; it has to first make percepts and then out of some of these, to make memories. The mystery about how percepts are made has been worked on for most of the 20th Century, and while not everything is understood, the basic problem has been pretty well solved. The brain makes percepts by using sensory codes that transform experience into patterns of brain cell activity within our visual, auditory, touch, taste, and smell systems. But the mystery of how the brain transforms some of these percepts into memories appears to be far more difficult. We believe that the brain has memory codes as well as sensory codes.

These are the codes for the importance of an experience; that is, for just how mentally significant an event is to us. Obviously, some events are more important than others. How does the brain separate them? We first need to review a little bit of basic brain function. We need to consider how the brain makes percepts, because it's out of percepts that memories come. Each of these topics will provide us with some clues about how the brain may make memories and reveal a potential memory code for the importance of an experience.

If we look inside the brain, we see brain cells. The brain is naturally made up of cells, specialized to receive and transmit information to other cells. They have to have special properties to do this.

A cell has three major parts. It has a cell body, which is triangular or parametly shaped. It has several stalks called dendrites, which receive information from other cells. It also has an output or an axon, which attempts to make a contact with a dendrite of another cell. There's a specialized place called a synaptic spine, an enormous network of billions and billions of cells with even more synapses.

The connections between cells are called synapses. The tip of an axon carries an electrical impulse called an action potential. When it gets to its end, it doesn't jump across to the next cell, but it releases little packets of chemicals called transmitters, which are held in the so-called vesicles at the end of the axon. And these will cross this little gap and attach themselves to a dendrite where they can cause a small electrical change. If these electrical changes are big enough, then they will cause an action potential or a series of action potentials to flow down the axon node of this recipient cell to other cells.

The synaptic cleft, is unimaginably small. To give you an idea, the size of a cleft in relation to us is about the same as the size of an average person compared to a person tall enough to touch the moon. At this very moment, millions of synapses are at work, sending these transmitter molecules across.

With a little basic brain function out of the way, we can now ask how the brain makes percepts. To being with, the brain has three major parts: a core, the most primitive and oldest in evolution, which explains our urge to run and hide, from the IRS, for example. A middle part to process emotions, explaining why we will pay a lot of money for popcorn when we know it isn't good for us; and the newest and largest part, the cerebral cortex. It is there where we think music, language, thought, and VCR instructions reside.

We've concentrated on the cerebral cortex because it is the largest part of the brain, and it certainly has been implicated in memories of a certain kind. To consider how the brain makes sensory percepts, we have to work on some sensory system, and we chose the auditory system many years ago. As sound enters the ear, it is transformed in the cochlea. Pressure waves hit our eardrum, and they are transformed inside of the cochlea into electrical impulses. Then they travel up pathways to the top of the brain, the auditory cortex of the human. The auditory system has various levels. We have worked on the auditory cortex of animals, and since the structure and the function of the cochlea are the same in all animals, we can use them as a model to understand the human auditory process.

Inside the auditory system are neurons, and they respond mainly to a small range of pitches of acoustic frequencies, from 1 to 13,000 kilohertz. The cell does not respond equally to all of these, but rather responds best to 6 kilohertz and less well to higher and lower frequencies. At 6 kilohertz it is at the peak of a function, called a Tuning Curve. The Tuning Curve is a range of frequencies a cell responds to. (Sometimes the Tuning Curve is called the Receptive Field, because it shows what this cell listens to in the whole auditory spectrum out there in the real world.)

How does the brain take an outside sound and package it and transform it in the brain into something that we can experience? It is by having cells that are selective in their response, in tune to a particular frequency. There are cells that are going to be able to respond to anything in the acoustic spectrum, which, for us, is quite broad, from about 50 cycles to about 25,000 cycles when you're young to about 15,000 cycles when your older. But there is not that much exciting stuff going on above 15,000 cycles anyway. A piano for instance only goes up to about 4,000 or 5,000.

Now having a sensory code, given by the Receptive Field or the Tuning Curve of a neuron, no cell in our brain responds to all sound. If it did, we would never have a clear perception of any single sound, because it would all be a big muddle. So we have a sensory code for acoustic frequency or pitch, coded by the preferential tuning of single cells to particular frequencies. Tuning Curves occur in all sensory systems. We're talking about the auditory system, but historically it has been believed that the tuning of neurons was fixed to be unchangeable either at birth or certainly within the first few years of development. An implication of this belief is that the adult brain is not very capable of tuning changes, and the adult brain is not very flexible. In particular, very mature brains, are quite limited. I disagree with the view that adult brains are not very plastic, capable of change, or even re-tuning, and I'll explain later some relevant findings.

At the neural systems level, events to the brain are sensory stimuli. They enter the cochlea where they're transduced or changed into electrical energy, which then goes into the lower auditory system and finally the upper auditory system. It is the discharges of cells in our auditory system that makes percepts. When the cells fire, we hear the sounds and perceive them. Engrams are entered into the process. Engrams are changes in the nervous system that underline memories. So how is sensory neural activity in the auditory system converted into engrams? This lasts a short time in the present or this can last forever.

We believe that this happens by the action of something called a Memory Code, which operates between the neural sensory activity in the auditory system or any sensory system, and the engram. The engram represents in some sense what that code is.

To have a learning situation in the laboratory, we used a very pervasive and basic form of learning called associative learning, when one event or stimulus regularly precedes another one, and then we associate the first with the second. The first becomes a signal for the second. If you take any arbitrary sound pattern and tell a young child the name of an object, they will learn that, make that association. The sound pattern is arbitrary, because dog is English for a certain animal, but it's a different word in different language. So we can associate the sound pattern "dog" with a real dog or we can associate "Rover" with a particular dog. Not only is all of our language based on associations, but much of our life is too. We learn that if one thing precedes another, we can use it as a predictor or signal. In the laboratory, we used something like a pure tone and follow that with something like food. And then the animal can learn that the tone is a signal for the food. We can then ask what happens to that tone in the brain when it becomes a signal? When it goes into memory.

Associations are very strong when the second stimulus is a reinforcer, something that we consider to be biologically or socially important. Some behavioral reinforcers are food, water, and sex. I don't think anyone would dispute that these are very important and that we tend to remember things that predict them. Love is also a reinforcer - affection, altruism, doing things for others without direct gain for you. Most people would not deny that money is a powerful reinforcer. Any behavior that avoids negative experiences, anything that will help you know when a negative experience is coming and how to avoid it certainly can be a reinforcer, as is the joy of knowledge and understanding. The important thing to remember is that the brain really is a curiosity machine, desiring information and drawing great satisfaction from learning about new things.

So now we know that a stimulus, or tone, is a signal for some reinforcer. Our task is to figure out how that tone goes from being a short lived percept into a long-term memory. We need to discover how the tone's importance as a signal is coded by the brain and where this information is stored. Prior to the time that we became involved in this particular problem, we and previous investigators had, in fact, discovered that learning that a sound that is important results in an increased response to that sound in the auditory cortex. When we perform an experiment to show how many times a cell responds to a tone, we set up a condition where the tone and the reinforcer are given randomly, so that the animal cannot use the tone as a predictor or as a signal. This gives us a baseline control condition. Then when the tone predicts the reinforcer and the animal learns that, there is a great increase in response of the cell. As we continue to have the animal learn more, the cell fires even more, showing that the auditory cortex increased it's response to an important stimulus.

This finding created a new mystery. What does this increased response really represent? The standard learning experiments always studied the response to a single tone, but we already knew that cells respond to many frequencies, and second, cells respond better, are tuned better, to some frequencies than to others. So, the mystery was what did this increased response really indicate about the brain's process of making percepts into memories? Findings about one frequency provide only a small part of the picture. Using only one tone to probe a cell's responses limits our view, our knowledge, and, therefore, our understanding of the brain's actions in storing information and making memories. We actually need to see the whole picture.

We decided that we would try to get a whole picture of the cell's responses by getting its tuning before and after to see if the animal had learned anything. First, we get the Tuning Curve by using many frequencies. We pick one of those frequencies, but not the peak of the Tuning Curve, and use that tone to signal a reinforcer. This can happen quite quickly, within a half-hour, even within five minutes. The animal's behavior can show that it has learned that the tone is a signal. After learning, we then obtain the Tuning Curve again and we compare the before learning and after curves to try to understand how the brain is processing information and representing the real world as a function of learning.

There are mainly two possibilities. The first is what most people had seen before us, that if you use a single tone as a signal, sometimes called a CS or Condition Signal, that from before training to after training there will be an increase in the number of discharges of the cell. However, that result can be very difficult to interpret. It's like looking through a crack in a window at midnight and trying to figure out what's outside. If you get the whole Tuning Curve, there are two possibilities. One possibility is that the increase at 8 kilohertz simply would be seen at all other frequencies, had anyone bothered to ask or interrogate the cell. This would be a general increase in response, and while interesting, something like the brain deciding to listen to everything better isn't going to reveal any specific information about 8 kilohertz and what it's meaning is. It's not very specific. But the same increase could represent a shift in tuning if there were decreases in response to some frequencies while there was an increase in response to the Condition Signal. It's analogous to the behavior of a radio when you turn up the volume, making everything louder. You don't get any specificity, and you don't get any information. It's just louder. This is like retuning the radio or the stereo to a new station. And so the question is since we already knew that responses to a Condition Signal would increase, was it a change that is general or was it a shift?

It's a shift.

This was the first time anyone had ever found that learning retunes the cells of a brain. This retuning affect is very pervasive and very strong. For example, occasionally you'll find cells that actually are tuned to two frequencies. But the effect is even stronger than that, because there are cells that may not be responsive to pure tones, but they can become responsive, and when they do they can become tuned to the training frequency. In our experiments, we found that before learning showed little response. When we trained the animal in this case to 10 kilohertz the tuning afterwards was a big, nice respond to 10 kilohertz.

Memories last a long time. The question is, really, how long do these memories or these tuning changes last? To put this in perspective, we need to consider the story of Bambi. All forest animals to this very day remember exactly where they were and what they were doing when they heard that Bambi's mother had been shot. For those of you who don't know the story, Bambi is the fawn, and hunters killed the mother. The moose says, "I was down by the edge of the lake at the time." And the possum says, "I was just about to cross the Interstate." The rabbit was in the glen, finishing a burrow. The bear was going to get crawdads, and the snake was shedding. So what does this story illustrate? It illustrates the fact that things that are very important to us tend to be remembered for a very, very long time. And so we have to ask if we make a tone important to an animal, does its effect last a long time? The answer is yes. When we trained a rodent at a higher frequency, at one hour, 24 hours, and one week, the cell was recorded and the tuning had shifted at one hour, and it stayed there 24 hours and it was still there a week later. So these tuning changes are not evanescent. In fact, they can be extremely long. We have tuning at one frequency, training at a higher frequency, and showing two weeks and four weeks after the training experience. We have taken this out as far as eight weeks or two months, which is a very long time in the life of a laboratory rodent. So these changes are very specific, where the retuning lasts a very long time. When animals learn that something's behaviorally important, actual cells in their auditory cortex shift their tuning. Now how could a cell become retuned?

Recall that the connections between neurons are called synapses, and it turns out synapses can be made either to become stronger or weaker. One possibility is that retuning involves making synapses for the signal stronger and making other synapses weaker. For example, on a scale where we have six frequencies, 1, 2, 4, 8, 16 and 32 kilohertz (before learning), the cell is tuned best to 4 kilohertz. The strongest synapses will make the cell give the most impulses, or 4 kilohertz in this case. When the animal was trained at 8 kilohertz, the tuning had shifted, a stronger synapse for 8 kilohertz. We actually have findings to support this kind of an explanation, that tuning shifts lead to an important prediction that if cells get tuned to a new frequency, then more of the area of the cerebral cortex ought to be tuned or specifically processing this new, important frequency. That specific prediction was tested and verified in another laboratory. Consider the work of Greg Rekanzone at the University of California San Francisco, who trained monkeys so that they would hear a tone of a certain frequency and they would get a fruit juice as a reward. After he had trained them to do this, he mapped their auditory cortexes, and he found the following. For animals that were tuned to 2.5 kilohertz, the amount of cortical area seemed best tuned to that training frequency compared to an untrained control frequency. For animals trained to 5 kilohertz, the cortex responded to that frequency compared to an untrained frequency. Similar findings were observed with two animals trained at 8 kilohertz. Therefore, it was determined that retuning results in an actual change in the allocation of neurons across the cortex.

Now let's get back to the fundamental issue of a potential memory code for storing the importance of experiences in the brain. What is this memory code that intervenes between sensory neural activity and the ultimate engram? That memory code, we believe, is the amount of tuning change in the brain. We've seen that the Receptive Field shifts from the original preferred frequency or best frequency to the frequency that the animal's trained at. Now it's our belief that as stimulus importance increases, so does the amount of tuning change in the brain, the number of cells that change their tuning. We think that stimulus importance actually has a memory code, and that a number of cells become tuned to that stimulus. We can now look at the big picture of both percepts and memories. And there's a lot on this, but it's really not very complicated. I already mentioned the psychological level of events, percepts, and memories. I mentioned this neural system's level of sensory stimuli into neural activity and engrams. At the level of single neurons, we find that there is electrical activity in the cochlea, but we have action potentials in response to a tone, and when those action potentials occur, we presumably experience that tone. Some of that information becomes an engram by a modification of synaptic strength.

Moving on, I'd like to talk about the mechanisms of making percepts into memories. And this is going to have clinical relevance as well as relevance for daily life. Imagine a simple diagram of the auditory system. We have the ear, we have the auditory system, and then we have the auditory cortex. Now can this explain how memories are made? The answer is no. There's nothing special in here that would tell the auditory cortex about the importance of anything. It would just give information about what we're hearing. But that would be a percept. That's that fleeting now. We have to remember that when there's a reinforcer, associations can become quite strong. So we have to ask a question of how does the reinforcer work with our sensory systems to make memory in the auditory cortex? There are some important clues about this.

The first point is that not only are there neural transmitters for messages that are sensory in nature, but there are neural transmitters whose job is not to bring information from the outside world, but rather to modify or modulate the actions of other transmitters. Then these are called Modulator Neural Transmitters, and probably the best studied and best known is acetylcohlene, or ACH. That's the first clue. The second clue is that ACH turns out to be important in learning, memory and cortical function and in fact, the two drugs that have been approved for treatment of Alzheimer's Disease, Eracept and Tacterin, are drugs that promote the actions of acetylcohlene in the brain. The third clue is that there is a structure in the base of the brain, the nucleus basalis, that provides acetylcohlene to the cortex. It's not a sensory structure or a motor structure. It's a structure that makes the cortex work well and can change it. And the fourth, and most important, clue is that if you record from cells in the nucleus basalis, those cells are activated by reinforcers. The nucleus basalis responds to reinforcers and in so doing, it releases acetylcohlene in the cortex.

So we hypothesized that if the brain normally makes memories of stimuli, signal reinforcers by the release of ACH, then we should be able to shift the cell's tuning without actually having the animal undergo a learning experience. So this is the idea. We have the auditory message coming up to the cortex, mediated through glutamate. We have the reinforcer causing nucleus basalis to release acetylcohlene. So here's the logic. Give a tone, but instead of giving a reinforcer, we have a tiny little electrode wire, it doesn't hurt anything, but it can be used to electrically activate the nucleus basalis. We pair the tone with stimulation of the nucleus basalis to see if we can change the tuning of the cells in the cortex just the same way that normal learning with a tone and a reinforcer would do.

Now if these findings were just confined to laboratory animals, they'd really be of limited interest. Why would we be interested in understanding the human brain and making it work better when it doesn't work well and helping it stay healthy and so forth? So the question is, does this all work the same way in the human brain? Obviously, people don't stimulate the nucleus basalis, but they can do other kinds of experiments. And last year, Morris, Fristen & Dolan of University College in London reported the following experiment in the proceedings of the Royal Society of London. They found a kind of a tuning change in the auditory cortex. These changes were positively related to changes of activity in the basal forebrain. These results agree with animals single unit data, and support neurobiological models of learning in the auditory cortex.

Now we are not just in the laboratory anymore. We're with real human beings, learning real things. So we feel that these data from the human studies are really quite important in suggesting that we have a basis for understanding how learning and memory formation take place in the human. The final model has three points. We call them "take home messages." The first is that learning actually changes tuning. Cells are returned to important sounds. Secondly, that tuning changes seem to be a memory code for stimulus importance. The more the tuning change, the greater the importance. And third, that these tuning changes, or the coding of stimulus importance in our brains, seemed to be caused by the release of acetylcohlene in the cortex.

I think that the findings are relevant both for normal and pathological situations.

But I think there is another implication, and this is one that deals with neuropathology, which is the fact that we seem to be able to retain important information, important memories, better than other memories after something bad happens to our brains, such as a stroke or some other damage.

We now know that new synapses can actually be formed and there's recent evidence that new cells can be born, even in adult brains, even in brains covered with gray hair. So I think that this is possibly the way things work. There's a bigger representation that we could learn about all these other things in the world and still have plenty of capacity in our brain. It doesn't have to be the case that if you remember something well, you can't learn anything else. Quite the contrary, there really seems to be no limitation on what a normal, healthy brain can do and learn, and there's no limitation with respect to age.