|
| |
The Dynamics of Human Memory
From: Cambridge University Press
| By:
Steven P.R. Rose |
EDITOR'S INTRODUCTION |
How much do we know about the workings of human memory? What are its abilities and limits? How has our understanding of the phenomenon evolved, and in what way is current scholarship changing our traditional conceptions? Steven Rose, professor of biology and the director of the Brain and Behaviour Research Group at the Open University, UK, reflects on some of these issues and, in the process, raises questions about, and offers insights into, a topic which has fascinated scientists, psychologists and philosophers for centuries. |
 t is not brains that make memories; it is people, who use their brains to do so. And animals, non-human animals, also make memories, and can learn and change their behaviour as a result of experience. Even some animals without much in the way of brains at all, just rather basic nervous systems, can do it. What this points to is the tremendous importance that the capacity to learn and remember has for the survival of animals. Plants do not need nervous systems, because all they have to do is to stand around with their arms--or branches--spread wide so that their leaves can catch the sun and photosynthesize. But animals which live on plants, and even more so animals which live on other animals, have to use their wits to find and capture their prey, and to avoid being eaten in their turn at least long enough to be able to reproduce. |
Such ways of making a living in the world demand the development of sensitive sense organs, and the capacity to register and interpret the data provided by those sense organs, to compare it with past experience and, even more, with the outcomes of that past experience. And this is what learning and memory are all about. It is not the only route to evolutionary success. After all, bacteria do pretty well without either brains or nervous systems, or even much by way of memory--though there have been some disputed claims that they can learn from experience. But once evolutionary strategies based on brains appeared, the pressure to get bigger and presumably smarter brains, with more capacity to learn and remember, must have been substantial. |
Memory and identity
To me, memory is the feature that defines every single one of us as an individual. We can contemplate losing a limb or a sense, or even having a heart or kidney transplant, and still retain a conviction, albeit modified, of our own personhood. Imagine losing memory--or having a memory transplant à la Manchurian Candidate--and the difference is immediate and apparent. We are our memories. In old age we can retain memories of our own childhood eighty or ninety years previously, even though every molecule in our bodies, and every cell except our neurons, has been recycled many thousands, perhaps millions, of times. In many ways the problem of memory is similar to that of the problem of form--how is it that biological form is retained despite this constant demolition and rebuilding of our body substance? And of course this is why the diseases of memory, such as Alzheimer's are so devastating. |
Neuroscience
It is in part for these reasons that as a neuroscientist, I have been fascinated by memory for most of my working life. I am constantly intrigued by the way in which a feature of our own existence which is so intensely personal, individual and subjective can none the less be explored by the "objective" methodology of science. The sense of confidence, even arrogance, amongst the industrial world's neuroscientific community is tangible. Around 25,000 researchers meet each year at the annual jamboree of the American Society for Neuroscience; our once recondite research field has become big business for pharmaceutical companies and putative genetic engineers. Bandwaggoners, from molecular biologists and quantum physicists to artificial intelligence (AI) modellers, are offering to tackle the Big Question, that of Consciousness. |
My aims are more modest. I have argued before and will repeat here that the study of learning and memory may be the key to deciphering the translation rules which lie between the languages of brain and mind, the Rosetta Stone of the neurosciences. There is a methodological principle operating here. In science it is always easier to study change than stasis. When a person or an animal learns, their behaviour changes in specific ways, and one can then ask what changes in the brain to match, to correlate with, the change in behaviour. This approach has lain at the heart of my own experimental programme over the past getting on thirty years. |
The limits to memory
But let us begin at the beginning, with the study of the limits of human memory itself. Fascination with the strengths and weaknesses of human memory go back to ancient times. Frances Yates, in her book The Art of Memory, described the mythic origins of the Greek and Roman techniques for cultivating memory. The poet Simonides is commissioned to sing the praises of his host at a Greek banquet, but makes the mistake of praising the twin gods Castor and Pollux at the same time. When the time comes for him to be paid, his host offers him only half, suggesting he get the rest from the gods. Just then, Simonides is summoned from the banqueting hall by two young men who turn out (surprise, surprise) to be Castor and Pollux themselves. In his absence the hall collapses, crushing host and guests beyond recognition. The grieving relatives ask Simonides to help to identify the corpses, and he is able to do so by recalling the order in which the feasters were sitting round the table. This became the basis on which Greeks and Romans cultivated what they called artificial memory. To recall (in the days before notepaper and autocues) a speech you were to give in the Senate, you decided on the points you wished to make and then "placed" them close to objects in a familiar environment so that you could recover them in sequence as you walked, in your imagination, through that environment. |
This technique, surprising though it may seem, works rather well. Yates traced it through the "memory theatres" (Figure 1) which culminated in the hermeneutic tradition of Giordano Bruno; but if even today you answer those advertisements in the Sunday newspapers which offer you methods for improving your memory, you will find yourself offered a similar method. The current holder of the record for remembering pi to 4000 places--or whatever the figure is--told me he used a similar technique, but believed he had invented it himself. And the Russian neuro-psychologist Alexander Luria, in his extraordinary little book The Mind of the Mnemonist, describes how one patient, a man made desperate by an incapacity to forget, remembered nonsense formulae Luria had given him over a period of practically two decades. Indeed Luria's case illustrates another point about memory--forgetting is functional. To remember all the data which pass through one's senses every day would be impossible; your memory would become, in the words of one of Jorge Luis Borges' memorable (sic) characters, "a garbage disposal." |
Graeco-Roman experiments with artificial memory also reveal another feature of the way in which the study of memory has proceeded over the centuries. Perhaps more than other sciences, biology has always worked with metaphor, and throughout history the brain and its properties have been analogized to whatever have been the best technological artefacts available to the culture. For the Greeks and Romans, memory was inscribed on wax tablets. Later, in the Renaissance, clockwork or hydraulic metaphors emerged. By the eighteenth and nineteenth centuries, the brain was a telegraph or telephone exchange. Today of course it is a computer, but I am going to suggest another way of modelling memory. |
Suppose I flashed a set of seven numbers on to a screen and asked you to repeat them back. You would probably do so without fail. How about eight numbers? It gets harder. Nine, and most people are lost. What? Unable to remember a sequence of eight digits? A simple calculation shows that to remember this requires a mere 41.86 bits of information. A basic pocket calculator has a 1000 bit memory, and the floppy disc of the Macintosh on which I am typing this more than ten million. So computers win hands down. Yet I can easily remember and read off to you a forty digit long string of numbers, and I am not Luria's patient or winning a place in the Guinness Book of Records. How do I, and others, do it? Simple: the numbers are sequences which I know as telephone numbers or birthdates. |
I don't view computer memory as a metaphor for human memory. As I see it, computers deal with dead, static information, locked into files, that can be pulled out, inspected and replaced. Humans deal not with information but with meaning. How many bits of information are there in my memory of my fourth birthday party, the colour of my grandson's hair or even what I had for breakfast this morning? Incomputable. Yet the AI people persist in confusing them, as when my fellow memory researcher at Oxford, Edmund Rolls, calculates that the primate hippocampus can store precisely (I like the precision) 36,500 distinct memories. |
I do not have to go all the way with the mathematician Roger Penrose into quantum speculation to dismiss the presumed relationship between computer memory and human memory as not much more informative than a wax tablet or hydraulic metaphor. Which is incidentally why I did not mind whether the computer Deep Blue did or did not beat Gary Kasparov at chess. Chess is a game of "pure" cognitive analysis. Human memory, human action, demands not just cognition but affect too--which is why it may be fun to play chess against a computer, but pretty boring to play poker. I would like to replace the famous Turing test for human-like computers with a poker test instead. I made the personal transition between chess and poker as an undergraduate in the late fifties, and I have never looked back. |
The taxonomy of memory
This business of remembering numbers illustrates some other features of human memory too. People forget a seven figure number quickly unless it is an important one such as a telephone number. There is a temporal gradient in memory--some things are remembered only for a short time, others for much longer. There is a great deal of evidence that the transition from short- to long-term memory is a crucial step, both in terms of function, molecular and cellular mechanisms, and what might go wrong in disease states. Short-term versus long-term memory is the first of the dichotomies in memory studies. |
This is another. Seven or eight numbers may be all you are able to recall, but suppose one tests your memory in a different way. Consider an experiment done by Lionel Standing, a Canadian psychologist, in the 1970s. He showed groups of subjects sequences of photographs, one after another, each for a few seconds. A week later, he called them in again and now showed them pairs of photographs--one they had seen before, and the other novel--and asked them to identify the one they had previously seen. His subjects could recognize up to 10,000 photographs with 90% accuracy. So recognition memory seems quite different from recall memory--and if that surprises you, think of the difficulty we all have in describing even a well-loved friend's face compared with the ease with which we recognize the person when we see them. We remember abstract information--such as numbers--differently from how we remember complex scenes or patterns. |
Inferring function from dysfunction
The taxonomy of human memory (Figure 2) owes much to the study of accidents of nature, brain lesions caused by stroke or damage or degenerative diseases such as Alzheimer's. The generally accepted taxonomy is due to the San Diego neuropsychologist Larry Squire, who distinguishes first between declarative and procedural memory--knowing that and knowing how. Knowing that an object with a saddle, two wheels and handlebars is called a bicycle is declarative; knowing how to get on it and ride it is procedural. These two types of memory seem to rely on different cellular, brain and bodily mechanisms, for the first is much more readily lost than the second (you can ride a bike even after a twenty year gap with almost no need for rehearsal). Declarative memory subdivides into semantic (knowing what doctors do) and episodic (knowing that I went to the doctor's last Tuesday). This latter is the most vulnerable in conditions like Alzheimer's. |
Accidents of nature provided until relatively recently virtually the only way to study the processes of human memory. In a recent book, Rewriting the Soul, the philosopher and statistician Ian Hacking dates the birth of the sciences of memory to the 1870s in France, with the recognition that there were specific diseases of memory. The study of recovery of memory after blows to the head--concussion--led to the recognition of the relationship between short- and long-term memory. It appears that, after such a blow, memory for events up to a few minutes prior to the accident can be recovered but these last moments appear to be irretrievably lost--hence the argument that memory must be "stored" in the brain in a more vulnerable form. Once past this critical time memory seems relatively invulnerable to such accidents.
 |
Perhaps the most famous and prolific exploitation of an iatrogenic accident in the history of medicine relates to a man known in the literature only by his initials, H.M. Operated on in the 1950s in Montreal for epilepsy, H.M.'s hippocampus (Figure 3) and portions of his temporal lobe were removed. The result was that H.M., whilst retaining his memories up until the time of the operation, has forever lost the capacity to transfer new memories from short- to long-term store. He remembers events only for a few minutes, before they fade from history. Every day begins afresh for him, back in a mental timewarp in which he is still a young man with the interests, attitudes and knowledge that he had before the operation was carried out. |
From this neuropsychologists concluded that the deep, neuron-rich region of the cerebrum, called the hippocampus for its fancied resemblance to a seahorse, was a structure essential for memory, or at least for the registration of short-term memory and its subsequent transfer to long-term store. The hippocampus has remained a controversial focus of physiological attention ever since. But a word of caution is required. Inferring function from dysfunction is notoriously difficult, even for inanimate objects. As Richard Gregory pointed out many years ago, if you remove a transistor from a radio and it emits a howl rather than a symphony, you cannot conclude that the function of the transistor is to be a howl-suppressor. You are studying not the absent part but the system in the absence of the part. And the dynamic plasticity of living systems, the multiply redundant structures and functions of the brain, make what is true even for transistor radios dramatically the case for brains. |
Normal memory in humans and animals
So we need to study functional as well as dysfunctional memory. The classic psychological studies of memory begin with Hermann Ebbinghaus, a German psychologist of the nineteenth century, who set himself to learn, and later to recall, lists of nonsense syllables. Such studies confirmed that the loss of memory occurs mainly in the first hour or so, after which what is retained is relatively stable. The study of normal human memory has remained a fertile field ever since, but in a strange way fails almost entirely to connect to the neurobiological literature. |
The most authoritative recent book on human memory, called just that, is by Alan Baddeley.  It contains virtually no references in common with the standard neurobiological books on memory. It concentrates on memory in everyday life, on working memory, on forgetting, on strategies for remembering, on so-called "flashbulb" memories and on autobiographical memory. None of these translates readily into the laboratory languages used in neurobiological circles. Despite all the claims for the convergence of neuroscience and psychology over the past decades we clearly still have a long way to go. |
Perhaps that is because almost at the same time as Ebbinghaus in Germany was founding the study of normal human memory, Ivan Pavlov in Russia was beginning to explore the physiology of animal memory. Not merely could his dogs learn--any shepherd could have told him that--but they could do so under standard, quantifiable laboratory conditions. The study of classical conditioning began with the famous dog salivating to the sound of the bell which heralded the arrival of its food. Researchers began to ask questions such as how long could the experimenter extend the interval between bell and food and still achieve the "association," could one extinguish the association by sounding the bell and not bringing the food, and so forth. |
And a few years later into the present century, along came J. B. Watson and B. F. Skinner and gave us behaviourism and operant conditioning, in which the animal learns to do something like press a lever or move to a particular site in its cage to obtain a "reward" of food or water, or to avoid "punishment" such as an electric shock. Skinner called these "contingencies of reinforcement" and offered to turn psychology into an exact science, a physics of behaviour, in which the brain was a black box whose internal content was irrelevant to the output of the animal in response to these varied contingencies. Pavlov and Skinner ruled unchallenged until the birth of modern neuroscience and cognitive psychology in the 1960s; today we are interested in these previous approaches mainly as historical blips; both approaches to understanding memory have proved scientifically sterile. |
Recent developments
It is literally only within the last decade that new methods to explore the functioning human brain have emerged, with the development of powerful and relatively non-invasive imaging techniques, such as PET (positron emission tomography) and MEG (magnetoencephalography). Neither is perfect: PET (Figure 4) gives good localization but not in real time, and because it involves radioactivity cannot be used repeatedly on the same subject; MEG (Figure 5), in its infancy as yet, gives superb time-resolution but poor spatial location. Both are, however, beginning to reveal insights into the dynamics of learning and memory. For instance, when a person is asked to learn and recall a word list, the initial learning processes activate regions of the left frontal cortex--that is, these regions show enhanced blood flow and glucose and oxygen utilization, taken as surrogate measures for neural activity. By contrast, when the subject is asked a brief time later to recall the same list, regions of the right hemisphere become engaged. The problem with all these measures at the moment is that, dramatic and beautiful as their resulting pictures may be, we do not know yet what they mean in terms of the brain structures, chemistry or the detailed functions involved. |
|
| |
