Your Memory: A User's Guide

Chapter 1
WHAT IS MEMORY?

"I have a terrible memory." How often have you heard this statement? In my own case, whenever I meet someone and in casual conversation admit that I carry out research on memory, by far the most common response is "You should do some work on me -- my memory is awful!" So is mine -- I once managed to forget to turn up for a radio phone-in show on memory. I was reminded of my lapse by reading the radio listings in the newspaper, and arrived at the studio just in time to be asked by the host for "a few tips on improving your memory"!

Yet I also believe that I have a good memory, and would argue, despite its occasionally embarrassing fallibility, that both my memory and yours exceed that of the best computer in terms of capacity, flexibility and durability. In the chapters that follow, I hope to persuade you to share my admiration.

Perhaps the best way to appreciate the importance of memory is to consider what it would be like to live without it, or rather without them, since memory is not a single organ like the heart or liver, but an alliance of systems that work together, allowing us to learn from the past and predict the future.

It doesn't take much to remind us of the frailty and impermanence of memory. Almost any damage to the brain will lead to some slow-down in learning and some impairment in the speed with which we access old memories. Certain areas of the brain, however, are particularly crucial for memory. Serious damage to these can lead to dense amnesia, which can be a crippling handicap.

Consider the case of Clive Wearing, a talented musician and an expert on early music, who fell ill as a result of a viral infection. Carriedby a large percentage of the population, the Herpes simplex virus typically has no worse effect than causing the occasional cold sore. On very rare occasions, however, the virus manages to overcome the brain's natural defenses and causes an inflammation known as encephalitis. This can lead to extensive brain damage, and until relatively recently was frequently fatal. Although the disease can now be treated, patients often suffer extensive brain damage which frequently leads to memory problems.

Clive Wearing is a particularly dramatic example of the terrible aftereffects of encephalitis. He is so impaired that he cannot remember what happened more than a few minutes before, with the result that he is convinced that he has only just recovered consciousness. He keeps a diary which records this obsession -- page upon page of records indicating the date, the time and the fact that consciousness has just been regained, when confronted with evidence of earlier apparent conscious awareness, by being shown a video of himself, for example, he becomes upset and denies the evidence, even after many years of being in this condition. It is as if, faced with the enormity of a life limited to a horizon of a few seconds, he clings to the view that he has just recovered consciousness, with the implication that in the future all will be well.

Clive's world was very effectively portrayed in a television program by Jonathan Miller entitled Prisoner of Consciousness. whenever his wife appears, Clive greets her with the joy appropriate to someone who has not seen a loved one for many months. She has only to leave the room for two or three minutes and return for the joy to be repeated, a process that is always full of emotion, and always expressed in the same way. Clive lives in a permanent present, unable to register change or to use the past to anticipate the future, a situation he once described as "Hell on earth. It's like being dead -- all the bloody time!"

Clive's memory for his past is less dramatically impaired than his ongoing memory. Nevertheless it is severely disrupted -- he knows who he is, and can give you a broad outline of his earlier life, but with very little accurate detail. He was not certain, for instance, whether his current, second, wife and he were married or not. He could remember, given appropriate cues, certain highlights of his life, such as singing for the Pope during a papal visit to London or directing the first performance of Handel's Messiah in London with authentic instruments and decor. He had written a book on the early composer Lassus, but could remember virtually nothing about him. His visual memory was also impaired -- he had spent four years in Cambridge, but did not recognize a photograph of his old college. His general knowledge was similarly reduced -- he had no idea, for example, who was the author of Romeo and Juliet.

There was, however, one area that was remarkably preserved, namely his musical skills. On one occasion his wife returned home to discover that his old choir was visiting him, and that he was conducting them just as he did in the old days. He could sight-read music and was able to accompany himself on the harpsichord, playing quite complex music and singing with great skill and feeling. Alas, he appears to find the transition from music back to his desolate state of amnesia particularly disturbing, with the result that music does not seem to provide the kind of solace that one might have hoped.

Clive has been in this state since 1985. He is still convinced that he has just woken up. He still lives in a desolate, eternal present. He cannot enjoy books because he cannot follow their plot, and takes no interest in current affairs because, likewise, they are meaningless as he-does not remember their context. If he goes out, he immediately becomes lost. He is indeed a prisoner limited to a brief island of consciousness in the sea of amnesia.

The tragic case of Clive Wearing demonstrates that memory is important, but what is memory?

The physical basis of memory

It is often assumed by non-psychologists, and indeed by a few psychologists, that psychological theories should have the final aim of giving a physiological account of psychological facts. This view, which is sometimes called reductionism, sees a continuous chain of explanation, extending down from psychology through physiology, biochemistry, biophysics and so on, right down to the subatomic particles studied by physicists.

Suppose I were an architect and wanted to find out about London's St. Paul's Cathedral. I could pursue my enquiries at many different levels. I could ask about the history of the building and how it came to be built following the Great Fire of 1666. I could ask about the style, and the influence of classical architecture on Sir Christopher Wren, who built it. I could ask about its function, and the details of the material which went into its construction. The notion that a study of memory must begin with its biochemistry would be somewhat analogous to advocating that anyone interested in St. Paul's Cathedral should begin by studying the atomic structure of brick and stone. While there is no doubt that such a study would be relevant (and indeed if the atomic structure of the bricks had been inappropriate, the cathedral would never have stood up), we could know everything about the atomic structure of brick and stone and yet know virtually nothing of interest about the cathedral. On the other hand, we could know a great deal about the cathedral without having any knowledge of the physiochemical properties of brick and stone.

The structure of materials does of course at some point constrain an architect and obviously has an important bearing on the creation of a building. Similarly, in principle, a number of aspects of human memory could be importantly influenced by physiological or biochemical findings. However, many of the claims for an understanding of the molecular basis of memory that were being made a few years ago have since been shown to be premature. The neurochemistry of memory is proving much more complex than was previously suspected. There is no doubt that progress is being made in this important area, and that one day there may be a very fruitful collaboration between the experimental psychologist and the neurochemist. Today, however, there is little area of overlap, so I will give only a brief account of some of the work concerned with the neurophysiology of learning and memory.

The neurophysiology of learning and memory

Learning almost certainly involves a chain of electrophysiological and neurochemical changes in the brain. Such changes are currently very difficult to study in the human brain, but considerable progress is being made in understanding the processes involved in learning in less complex organisms. For example, Eric Kandel has worked on the very simple marine organism Aplysia, which combines neuronal simplicity with a capacity for simple learning. It is capable, for example, of showing the phenomenon known as habituation. This is a process whereby a stimulus that initially evokes a response gradually comes to be ignored when it is repeated, in the absence of any positive or negative outcome. In the case of Aplysia, if its siphon is stimulated, both the siphon and the gill tend to be withdrawn initially; after repeated stimulation the withdrawal response stops, an effect that can last from minutes up to weeks. The withdrawal response involves electrical transmission across synapses, the special junctions between neurons, or nerve cells. Transmission across synapses depends on neurotransmitters, chemical messengers that allow one neuron to communicate with another. These in turn depend upon the release of calcium ions. The process of repeated stimulation gradually reduces the activity of the channels that release calcium ions, thus reducing the likelihood that sufficient calcium ions will be released to cause firing or onward transmission of a nerve impulse.

The opposite of habituation is sensitization, a process that occurs when an independent stimulus increases the probability of a response. Hearing a shot, for example, might make you sufficiently jumpy to be startled by the sound of a car door slamming subsequently. In the case of Aplysia, an unpleasant stimulus to the tail enhances the withdrawal response when the siphon is touched. This is caused by an increase in the amount of neurotransmitter substance released as a result of a greater influx of calcium ions into the terminal part of the neuron.

Aplysia is also capable of the form of learning known as classical conditioning. The best known example of classical conditioning was that observed in dogs by the Russian physiologist Ivan Pavlov, who showed that when the presentation of food was regularly associated with a bell, eventually the sound of the bell alone led to salivation. In the case of Aplysia, the equivalent to food is a strong stimulation to the tail, which causes the automatic response of withdrawing the gill and siphon. The equivalent of the bell is a mild touch of the siphon, which does not of itself lead to withdrawal. However, when the light touch is consistently followed by a strong tail stimulus, it eventually leads to withdrawal of the gill and siphon in the absence of the tail stimulus. This simple. analog of learning can persist for several days. Kandel suggests that the underlying mechanism is similar to that of sensitization; the light touch to the siphon eventually leads, through association with the stronger tail stimulus, to an increase in the flow of calcium ions into the terminal part of the neuron, leading to firing and transmission of the nerve impulse across the synapse.

The underlying mechanism for the more complex aspects of learning and memory remains in doubt. However, one possible mechanism is suggested by the effect known as long-term potentiation (LTP), a phenomenon first discovered by Timothy Bliss and Terge Lomo in the 1970s. while working on the hippocampus (assumed to be crucially involved in learning and memory) in rabbits, they found that intense electrical stimulation of connected areas resulted in hippocampal cells responding more strongly to stimuli than they had done previously. This enhanced response lasted for days, weeks, and even longer, suggesting that it might be a mechanism for long-term learning.

Subsequent research has indicated that LTP depends upon the activity of the receptors on both sides of the synapse. When the pre-synaptic sending mechanism receives high-frequency stimulation, it releases the neurotransmitter glutamate. For LTP to occur, however, the post-synaptic or receiving neuron must also be operating at the appropriate level. The relevant post-synaptic receptors are sensitive to a substance known as N-Methyl-D-Aspartate (NMDA), and firing depends on having exactly the right balance of ions in the receptor channel. When both pre- and post-synaptic circumstances are right, the nature of the synapse changes, so that in future a much weaker pre-synaptic stimulus will cause the post-synaptic neuron to fire.

The fact that the cells associated with LTP are particularly numerous in the hippocampus provides some encouragement for believing that this may indeed be a basic learning mechanism. In a classic book published in 1949, The Organization of Behavior, the Canadian psychologist Donald Hebb speculated that a mechanism such as this might underlie the process of learning. Since that time a number of computer-based learning models have been developed using Hebb's ideas.

How psychologists study memory

While some psychologists are involved in trying to understand the physiological basis of memory, this is not the most common approach and will play a minor role in the remainder of this account of human memory. If psychologists do not study memory by examining its physical or biochemical characteristics, how do they arrive at their findings? Do they simply ask people how they remember things? On the whole, no. While it is unwise to ignore people's comments on how they learn or remember, experience has shown that this kind of information is an unreliable source of evidence.

Consider, for example, the question of visual imagery. Nineteenth century British scientist Sir Francis Galton did a classic study which involved writing to a large number of eminent men and asking them to try to conjure up an image of their breakfast tables on the morning they received this unusual request. They were asked to comment at length on the richness, detail and vividness of the image they created, and enormously wide differences were observed, with some respondents reporting that their remembered breakfast table was almost as vivid as their direct perception of it, while others reported no imagery at all. Subsequent work has confirmed that people differ extremely in the reported vividness of their imagery. Yet attempts to relate this to their memory abilities have proved universally disappointing. For example, Sir Frederick Bartlett had his subjects try to recall stories, and noted that although those who claimed to have vivid visual imagery were on the whole more confident in their powers of memory than those without such imagery, they were no more accurate in their recall. A much later study by three American investigators, di Vesta, Ingersoll and Sunshine, looked at the relationship between stated vividness of imagery and a range of other tests. They found memory performance was not related to vividness of imagery; indeed the only measure that did show relationship to imagery was a measure called "social desirability," an indicator of the extent to which subjects attempt to be obliging and give socially acceptable answers! Thus while large differences in the reported use of visual imagery exist, they do not seem to tell us very much about the functioning of human memory, whereas other methods based on performance rather than self-exam have proved very fruitful.

If people's comments on their own memory are unreliable, how do we investigate memory? The answer is by setting subjects various memory tasks and scoring how well or badly they do them. Sometimes experiments take advantage of participants' differential memory abilities, but more frequently they take advantage of the difficulties people have and the mistakes they make when asked to remember certain types of material. If I were to present you with a string of consonants, say l r p f q h, and ask you to repeat them back to me, you would probably get most of them right, but your occasional errors would be revealing; for example, you would tend to substitute b for p or s for f, the errors being similar in sound to the correct item. I would conclude from this, as Conrad and Hull did, that you used verbal or acoustic memory rather than visual memory in order to remember the letters.

Another way of exploring human memory is to use a method known as "selective interference." might, for example, want to test the idea that people remember addresses or telephone numbers by repeating them under their breath. I could prevent such repetition and see if it impairs recall. Or I could ask someone to articulate an irrelevant word such as "the" while they are trying to rehearse or write down a telephone number and see if their performance drops dramatically.

The chapters that follow are concerned with human memory for a wide range of material, but you will no doubt notice an emphasis on verbal memory. The reasons for this are twofold. First, there is no doubt that verbal coding plays an extremely important part in human memory. Even when we are remembering visually presented items, or recalling actions or incidents, there is a strong tendency to supplement other aspects of memory by verbalizing, turning what may be initially purely a visual task into a combined visual and verbal one. The second reason for a predominance of verbal material is more practical. On the whole it is much easier to select and control verbal material than it is to manipulate visual, tactile or auditory stimuli. Suppose, for example, we want to study the effects of the familiarity of the material we are using. Information exists on the frequency with which every word in the English language is used, allowing us to easily quantify the familiarity variable. Similarly, data exist on the age at which people tend to first encounter particular words, on the tendency of a word to evoke a visual image, and so on, making verbal material by far the easiest to manipulate in experimental settings.

Another advantage of using words and letters as test materials is that they can be presented in the spoken or written mode, and can be recalled in either. With visual material, however, we are limited to one mode of presentation, and typically to testing by recognition, since it is hard for a subject to indicate visual recall other than by drawing, which has major limitations unless you're a talented artist.

As will become clear from the chapters that follow, psychologists investigating memory are largely in the position of someone trying to understand the functioning of a machine without being able to look inside it. Consequently they have to rely on manipulating the tasks that the machine must carry out, and on carefully observing its behavior under various conditions. Such an approach demands considerable patience and ingenuity but, as I hope you will agree by the end of this book, it can produce important insights.

The nature of human memory

While the plight of dive Wearing argues strongly for the general importance of memory, it does not tell us much about the detailed nature of the systems underlying human memory. Suppose we wanted to replace his faulty memory, what characteristics should our memory prosthesis provide?

Another way of asking the same question would be to take an evolutionary perspective and speculate on what memory functions might prove useful to an organism evolving in a complex and varied, but nevertheless structured, world. Let us assume that the organism has been given a number of sensory channels -- vision, hearing, touch and smell, for example. Information from these various channels should, in principle, be related; objects such as trees can be seen and touched, and indeed heard as the wind rustles through their leaves. Appreciating this and creating some representation of an object is likely to require memory, at least of a temporary form, a short-term or working memory that will allow the organism to pull together information from a number of sources and integrate it into a coherent view of the surrounding world.

It would also be useful to build up, over time, some knowledge about the world. Given that the world is at least partly predictable, it would be advantageous to learn which foods are good and which cause illness, for example. In short, some form of long-term memory would also be useful. Such long-term learning can be of several different kinds, however, and each kind seems to obey different rules. Clive Wearing retains his skills as a musical performer, but his capacity to retrieve facts about the past (details of his musical achievements, the names of great composers) is severely impaired.

Clues as to the structure of the complex alliance of systems that we call human memory are provided by other individuals with less dramatic memory problems, and of course by the study of the memory processes of normal people, as will become clear in later chapters. However, it might be helpful at this point to give a brief overview of the probable psychological structure of human memory, to provide a general framework within which the rest of the book can be interpreted.

The realization that memory can be fragmented into subcomponents is not a new one; it was proposed in the 1890s by the great American psychologist William James, and again by Donald Hebb in 1949. Experimental evidence for the fractionation of human memory has developed principally over the last 30 years. Until the 1960s many psychologists felt that it was unnecessary to assume more than one kind of memory, but by the early 1970s some form of distinction between long-and short-term memory was widely accepted. By the end of the decade both short- and long-term memory systems had been subdivided further.While not everyone would entirely agree with the structure I am going to propose, there is by now broad agreement that some such fractionation is useful. While I like to refer to separate systems and subsystems, other theorists might prefer to emphasize the different processes involved in remembering, rather than the underlying structures within which such processes operate. However we are likely to agree on virtually all of the information about human memory that I shall be presenting in this book. If there is disagreement, it is most likely to occur in areas where the evidence is too thin to allow us to decide between a number of plausible alternatives.

Consider, for example, Clive Wearing's inability to remember what he has had for breakfast. Why does he have this problem? One possibility is that the experience of having breakfast never registers in his brain; in other words no memory trace is laid down. A second possibility is that a trace is laid down, but fades away very rapidly. A third possibility is that the memory trace is there, but cannot be accessed or retrieved. The memory trace may be like a book in a library with no catalogue system. As we shall see later, deciding which, if any, of these scenarios is responsible for an observed memory failure is extremely difficult. Nevertheless it is important and potentially helpful to bear in mind that any adequate memory system must be capable of registering information presented, storing that information over time, and retrieving it when required.

How many kinds of memory?

Intense controversy during the 1960s led to a whole range of memory models of a broadly similar form. They tended to assume three kinds of memory -- sensory memory, short-term memory and long-term memory -- and are well represented by the model proposed by Richard Atkinson and Richard Shiffrin. Because it was both typical and influential, this model acquired the nickname the modal model. In this model it is assumed that information comes in from the environment through a parallel series of brief sensory memory stores and goes into a common short-term store. This is assumed to act as a working memory, capable of manipulating information and relating it to long-term storage. Indeed the short-term store forms a crucial link in this model; without it, neither the learning of new material nor the recollection of old information is possible. We will consider each of these subcomponents separately.

Sensory memory

When you go to the movies you see what appears to be a continuous scene in which people apparently move quite normally. What is actually presented to your eyes is a series of frozen images interspersed with brief periods of darkness. In order to see a continuously moving image it is necessary for the brain's visual system to store the information from one frame until the arrival of the next. The visual store responsible for this is one of a whole series of sensory memory systems that are intimately involved in our perception of the world.

Even within visual memory there are probably many components capable of storing visual information for a brief period of time. If you move the end of a brightly glowing cigarette in a darkened room you will find that a trace is left behind -- you can write a letter of the alphabet and someone else will "see" the letter. This effect was used to measure the duration of the visual sensory memory trace as long ago as 1740 by a Swedish investigator, Johann Segner, who attached a glowing ember to a rotating wheel. When the wheel was rotated rapidly, a complete circle could be seen, since the trace left at the beginning of the circle was still glowing brightly by the time the ember returned to its starting point. If the wheel was moved slowly, only a partial circle would be seen, since the trace of the first part had faded by the time the ember returned to its starting point. By rotating the wheel at a speed which just allowed a complete circle to be drawn, and measuring the time taken for one revolution, Segner was able to estimate the duration of this brief sensory store. He found it to be approximately one-tenth of a second.

This phenomenon, known as "persistence of vision," can be demonstrated even more simply. Spread out the fingers of your hand and pass them in front of your eyes. Do so slowly at first and you will notice that the background seems unstable and tends to jump about. Now move your fingers back and forth rapidly. You will then see what appears to be the normal background, although a little blurred. With rapid movement, the scene is interrupted only briefly, allowing the information registered by the retina to be refreshed before it fades away.

There are at least two, and probably more, components to sensory visual memory, or iconic memory as it is sometimes called. One of these appears to depend on the retina of the eye and is primarily influenced by the brightness of the stimulus presented. The second one occurs at a point in the brain after information from both retinas has been received and integrated. This component is much more sensitive to pattern than to brightness, and represents the operation of a system involved in shape recognition.

An analogous series of sensory memories occurs in hearing. If I were to present an extremely brief click in one corner of the room, you would be very good at deciding from which direction the click came. In order to do this, you would use the tiny difference in the time of the click's arrival at your two ears, performing a task analogous to the use of sonar to locate the position of a ship. However, in order to make use of this discrepancy in time of the click's arrival at your two ears, it is necessary to have a system that will store the first click until the arrival of the second, allowing this difference to be estimated extremely accurately. While we would not term this a memory system in the usual sense, it certainly is a system for storing and retrieving information, and as such can legitimately be described as a very brief sensory memory system.

The existence of a more durable auditory memory system can be shown as follows. Suppose I were to read out to you a series of nine-digit telephone numbers. The chances are that you would get most of the digits of each number right, but would tend to make errors. If I then switched to a system of presenting the numbers visually, one digit at a time, you would discover that you would make rather more errors, particularly towards the end of the sequence. The graph above shows a typical error pattern for nine-digit sequences both read and heard.

The most striking feature of this graph is the discrepancy between the two modes of presentation in the case of the last item presented. When this is spoken, it is almost always correctly recalled; when it is presented visually, errors are numerous. The reason for this appears to be that when the sequence is spoken, the last item can still be recovered from a brief auditory memory, sometimes referred to as echoic memory since it is rather like an echo lingering on after the item has been spoken. The echo is limited to one or possibly two items. Consequently it can be wiped out by presenting a further irrelevant item afterwards. Echoic memory is left holding the irrelevant item instead of the last digit. Thus if I had spoken the sequence of digits to you, and then followed it with the spoken instruction "Recall," the "echo" of the last digit would have disappeared. The system involved in echoic memory of this type seems to be particularly geared to speech, since a simple but meaningless spoken sound such as "bah" will disrupt performance whereas a pure tone of similar loudness and duration will not. A sequence of spoken digits is better remembered than a sequence of digits presented visually because auditory sensory memory appears to be more durable than visual.

Auditory sensory memory is not limited to speech sounds. Suppose you are worried about some component in your car engine and you listen to it while driving. What you will be trying to hear is a repeated sound embedded in the relatively random engine noise. In order to perceive the repetition you need to be able to store a long enough "bite" of noise to be able to detect the one feature that seems to be recurring.

This effect has been used to study auditory memory. The listener is presented with a tape which recycles a sample of randomly fluctuating noise and the size of the sample is then systematically varied, If the sample is half a second long, the listener is required to perceive features that recur every half second. To be able to do this, he or she needs an auditory memory system that stores at least half a second's worth of sound. If the sample lasts for a full second, a more durable memory store would be needed to detect the rhythmic fluctuation. When faced with this task, subjects vary somewhat in their capabilities, but on average can detect repetitions separated by up to three seconds, indicating an auditory memory system of at least this duration.

Although we have touched only briefly on sensory memory, we shall not be returning to it. While it is an important component of our overall memory apparatus, it is probably best seen as part of the process of perception. To explore it further would demand far more detailed analysis of perception than is possible within the limits of the present book.

Short-term memory

To understand this sentence, you need to remember the beginning until you get to the end. Without some kind of memory for the words in it and the order in which they occur, it would be incomprehensible.

Suppose I ask you to multiply 23 by 7 in your head. Try looking away from the page and doing this. First of all, you need to remember the numbers involved. Then you will probably multiply 3 by 7, and remember that the answer is 21. Then you will remember the 1 and carry the 2. Then you will multiply the 2 by 7 and retrieve the 2 you carried, making 16. Then you will retrieve the original 1 and come up with the answer 161. All of this involves a good deal of temporary storage of numbers, all of which need to be retrieved accurately and at the appropriate time. Having completed the calculation, there is no further need to retrieve information such as which number was carried, and after a couple of similar calculations you will likely not be able to remember this information.

In both language comprehension and arithmetic, therefore, there is a need for the temporary storage of information in order to perform various functions subsidiary to understanding or calculating. Once the task has been achieved, the subsidiary information is no longer required. Short-term or working memory is the name given to this system, or perhaps more appropriately, set of systems. Information which is essential for a brief period of time is temporarily stored, then becomes irrelevant.

To what extent is short-term memory a system different from long-term memory? Here again there has been considerable controversy in recent years. One view is that short-term memory represents the same system as long-term memory, but is used under special conditions which lead to very little long-term retention. The alternative view, which I myself support, is that long- and short-term memory involve separate systems, although they are very closely integrated in operation. I myself would further argue that short-term memory represents not one but a complex set of interacting subsystems which I shall refer to as working memory.

Long-term memory

Of the three types of memory -- sensory memory, working memory and long-term memory -- the one that corresponds most closely to the lay person's view of memory is long-term memory. This represents information that is stored for considerable periods of time. Indeed, as we shall see later, some theorists claim that information in memory never disappears, but simply becomes less and less accessible. Remembering your own name, how to speak, where you lived as a child, or where you were last year or indeed five minutes ago are all assumed to depend on long-term memory. Such memory is primarily concerned with storing information, unlike sensory memory and short-term memory where the storage is an incidental feature of other aspects of the system.

To an experimental psychologist the phrase "long-term memory" refers to information which is stored sufficiently durably to be accessible over a period of anything more than a few seconds. The reason for this is that, on the whole, memory tested after one or two minutes seems to behave in much the same way as memory tested after one or two days, or years. The same does not apply, as we shall see, to memory tested after one or two seconds, or even milliseconds. Is long-term memory a unitary system? This is still a controversial question. Distinctions of at least two types are commonly made, however.

Episodic and semantic Long-term memory

A few years ago Canadian psychologist Endel Tulving made a useful distinction between two types of long-term memory: episodic memory, which involves remembering particular incidents, such as going to the dentist a week ago, and semantic memory, which essentially concerns knowledge about the world. Knowing the meaning of a word or the chemical formula for salt or the capital of France would all be examples of semantic memory. There is no doubt that there are differences between specific personal memories of individual incidents and generalized knowledge of the world, which has often been acquired over a considerable period of time. Whether these represent separate memory systems or different aspects of a single system is still uncertain. However, the distinction is a convenient and useful one. In this book semantic memory has a chapter of its own.

A great deal of research on episodic and semantic long-term memory has used verbal materials, since words are easy to present and people's responses are easy to record and score. In recent years researchers have increasingly asked whether memory for verbal materials is characteristic of all memory, and in particular whether memory for non-verbalizable sensory experiences relies on quite different memory systems. Undoubtedly we can remember the taste of cheese or the smell of burning rubber or the sound of the sea breaking on a rocky shore without using verbal descriptions of these experiences. Are there separate auditory and visual memory systems, or an all-embracing memory system which is capable of encoding all our experiences? Taking this latter view, much verbal learning is verbal only inasmuch as the material is presented verbally and subjects respond verbally; what is stored is the experience conjured up by the verbal material. Fortunately the general rules which apply to the learning of verbal material also seem to apply, at least broadly, to remembering pictures or sounds, so the overall conclusions drawn in the chapters that follow are still likely to be valid whether we conclude that long-term memory is a unitary, dual or multiple system.

Implicit and explicit memory

It has been known for many years that densely amnesic patients such as Clive Wearing may still be capable of certain kinds of new long-term learning. The learning of motor skills such as typing is typically preserved, as is a whole range of phenomena known as priming. This term refers to the observation that when a word or object is seen or heard more than once, it will be seen or heard more readily on second and later occasions. Thus if you have recently read the word rabbit, then you will be better able to perceive it if I present it very briefly, and will be more likely to come up with the word if asked to produce something that will fit the pattern of letters R-B--T, than a subject who has seen a quite different word.

Learning measured in this way is called implicit. Because the subject is not asked about earlier presentations of material being learned, their influence is reflected indirectly in the speed or nature of subsequent performance, typically in a non-memory task. Such learning is not affected by many of the factors which are important when learning is measured by recall or recognition. Processing a word in terms of its meaning, for example, enhances subsequent recall, but does not influence the magnitude of priming, whereas changing the physical presentation of the word (changing the typeface in which it is printed, for example) tends to reduce priming, but has little or no effect on recall.

Another form of implicit learning is conditioning, which, as we saw earlier, is possible even in rather simple organisms such as Aplysia. An amnesic patient will show a normal capacity to learn that a tone is followed by a brief puff of air to the eye, and will blink on hearing the tone, but will be unable to recollect the earlier training, or indeed explain the function of the nozzle that has repeatedly delivered the puff of air to his or her eye.

Amnesic patients are also capable of showing adaptation, another simple form of learning that occurs following repeated presentation of material. For example, we tend to respond negatively to unfamiliar stimuli, such as hearing a Korean melody for the first time. After several presentations, however, the melody will be judged as much more pleasant, not only by healthy subjects, but also by amnesic patients, who will nevertheless deny ever having heard such a melody before.

In addition to being preserved in amnesic patients, these various types of learning have a number of other features in common, for example, they are relatively insensitive to how much attention is paid during learning, or whether an attempt is made to process the new material in terms of its meaning. Both of these have a powerful effect on explicit learning. It seems unlikely, however, that the whole range of types of implicit learning which includes acquiring manual skills, priming, conditioning and adaptation, all reflect a single memory system. They typically rely on different parts of the brain, with different implicit learning processes being impaired in different types of patient. What they have in common is that they all represent relatively automatic kinds of learning which do not depend on your capacity to remember in the sense of recollecting past experience. This type of memory is severely impaired in amnesic patients such as Clive, and, as we shall see, is central to what we normally regard as our memory.