Memory, p.32
Memory, page 32
Joseph Farley, working with me at the Marine Biological Laboratory in Woods Hole, provided additional support for the type B cell’s memory record by inserting microelectrodes into B cells of living snails that had not been trained with light and rotation. By injecting electrical current shortly after a light flash occurred, he artificially enhanced the B cell responses and thereby mimicked the effect of learning on the type B cells. After allowing the animals to recover, he found that they showed evidence of learning the light-rotation association just as if they had been conditioned with natural stimuli.
We were, therefore, at a storage site. At this site the memory record ‘looked like’ more signals in response to light. More electrical signals send more synaptic messages to other neurons and therefore store the memory. But what in the neuron makes more electrical signals? We knew that the flow of charged particles through channels in the neuron’s wall produced the signals. Perhaps learning regulated channels? Memory-regulated channels would in turn alter the charged particle flow, the signals, and the synaptic messages.
We had to look at the channels. This would require another level of acrobatics with microelectrodes. One microelectrode in the type B neuron would not be enough. I now inserted two or three into the same cell. Inserting each microelectrode without knocking the others out or tearing the cell wall apart proved to be a nerve-racking balancing act. Yet I eventually managed to record this way for hours on end, sometimes with one of the microelectrodes within the wirelike axon, only a millionth of a meter in diameter. The signals recorded by such microelectrode ensembles revealed the flow of charged particles through the channels in the neuron’s wall.
Soon after I began measuring particle flow across the neuron wall, or membrane, something peculiar caught my attention. Once activated by the light flash, the flow of certain charged particles, those of a potassium salt, across the membrane could not be entirely reactivated for many seconds afterward. It was as if the potassium flow had a memory of its prior activation. I remember sitting in the dark marveling at the slowness of this recovery (originally observed by John Connor in other neurons) and thinking how different this was from the classic story worked out for electrical signals, which spread down axons at fantastic speeds. Each of these axonal signals, also present in the type B axon, was completed within a few thousandths of a second and ready to be reactivated almost immediately thereafter. Maybe, I thought, the prolonged recovery of the potassium flow was in some way related to the prolonged nature of memory storage. Maybe a recovery lasting many seconds could, during memory acquisition, be extended to many hours, days, and longer. When all was said and done, this would prove to be the case. A permanently altered flow of potassium particles through membrane channels provides a memory record for later recall not only in the snail, but also in mammals.
Now the words memory record and memory trace took on a new and exciting meaning. A memory record ‘looks like’ altered molecular channels in neuronal membranes. Channels in a mature neuron’s membrane remained changed for weeks after the light occurred together in time with rotation. Nothing comparable had ever been encountered in any fully formed cell known to science. This was not what I or anyone else had expected. Since Helmholtz first measured its speed, Bernstein theorized how it was generated, and Hodgkin and Huxley measured its underlying flow of charged particles, no one imagined that a neuron’s electrical signal could remain transformed for days or even weeks. True, the signal developed along with the neuron. But once development is complete, a neuron’s signals were thought to be just as constant as those that pace the heart. Yet our observations in the snail’s brain and later the rabbit’s hippocampus violated our expectations and led us to a new way of thinking about particle flow. Experience could produce long-lasting changes of particle flow. Nature heard nurture’s voice in the movements of particles through membrane channels …
The potassium channel changes that stored the memory of the light-rotation link in the snail were telling us, then, not to look for an increased number of synapses or greater complexity of branches but for modified membranes distributed throughout systems of neurons that already existed – that had already developed according to genetic blueprints. They were telling us that the neuronal code for memory has its own unique features, quite distinct from codes that allow for development of the neurons themselves. The initial changes during snail and, later, rabbit learning were saying that the cellular expressions of remembered time that are due to an individual animal’s experience are different from expressions of evolutionary time that are due to an entire species’ experience. And to understand memory, particularly memory as complex as that of humans, these differences would have to be sorted out.
From STEVEN PINKER, The Language Instinct (1994)
The most sophisticated estimate [of human vocabulary] comes from the psychologists William Nagy and Richard Anderson. They began with a list of 227,553 different words. Of these, 45,453 were simple roots and stems. Of the remaining 182,100 derivatives and compounds, they estimated that all but 42,080 could be understood in context by someone who knew their components. Thus there were a total of 44,453 + 42,080 = 88,533 listeme words. By sampling from this list and testing the sample, Nagy and Anderson estimated that an average American high school graduate knows 45,000 words – three times as many as Shakespeare managed to use! Actually, this is an underestimate, because proper names, numbers, foreign words, acronyms, and many common undecomposable compounds were excluded. There is no need to follow the rules of Scrabble in estimating vocabulary size; these forms are all listemes, and a person should be given credit for them. If they had been included, the average high school graduate would probably be credited with something like 60,000 words (a tetrabard?), and superior students, because they read more, would probably merit a figure twice as high, an octobard.
Is 60,000 words a lot or a little? It helps to think of how quickly they must have been learned. Word learning generally begins around the age of twelve months. Therefore, high school graduates, who have been at it for about seventeen years, must have been learning an average of ten new words a day continuously since their first birthdays, or about a new word every ninety waking minutes. Using similar techniques, we can estimate that an average six-year-old commands about 13,000 words (notwithstanding those dull, dull Dick and Jane reading primers, which were based on ridiculously lowball estimates). A bit of arithmetic shows that preliterate children, who are limited to ambient speech, must be lexical vacuum cleaners, inhaling a new word every two waking hours, day in, day out. Remember that we are talking about listemes, each involving an arbitrary pairing. Think about having to memorize a new batting average or treaty date or phone number every ninety minutes of your waking life since you took your first steps. The brain seems to be reserving an especially capacious storage space and an especially rapid transcribing mechanism for the mental dictionary. Indeed, naturalistic studies by the psychologist Susan Carey have shown that if you casually slip a new color word like olive into a conversation with a three-year-old, the child will probably remember something about it five weeks later.
From FRANCIS CRICK, The Astonishing Hypothesis: The Scientific Search for the Soul (1994)
What about short-term memory? What is known about that? A memory might be defined as a change in a system, due to experience, that makes some alteration to its subsequent thoughts or behaviour, but this is too broad to be of much value. It would cover fatigue, injury, poisoning, and so on, and would not distinguish between learning and development (early growth). The Israeli neurobiologist Yadin Dudai has produced a more useful and more sophisticated definition. He first describes what he means by an ‘internal representation’ of the ‘world’ – that is, of both the external and internal milieu. He defines an internal representation as ‘neuronally encoded, structured, versions of the world that could potentially guide behaviour’. This emphasises that, at bottom, we are mainly concerned with how nerve cells (neurons) influence behaviour. ‘Learning’ is then the creation or modification of such an internal representation, produced by experience. Such changes persist for an appreciable time (sometimes for years), although we shall be interested mainly in memories that last only a very short time.
I shall not be concerned with very simple forms of memory, such as habituation or sensitisation. (Suppose you show a baby the same picture ten times in succession. At first he is interested, but soon he becomes bored with it. This is called ‘habituation’.) These processes are classed as ‘nonassociative’. They occur even in very lowly animals, such as the sea slug. We shall be more interested in ‘associative learning’ in which the organism responds to relations among stimuli and actions.
It is useful to divide memory into several fairly distinct types, although exactly how they should be described is a matter of controversy. One convenient division is into episodic, categorical, and procedural memory. Episodic memory is a memory of an event, often together with irrelevant details associated with the event. A good example would be remembering where you were when you heard that President Kennedy had been assassinated. An example of a categorical memory would be the meaning of a word, such as ‘assassination’ or ‘dog’, whereas the knowledge of how to swim or to drive a car would be classed as procedural memory.
Another method of classification depends on timing: how long it takes to acquire the memory and how long the memory usually lasts. Some memories, especially episodic memories, are classed as ‘one-shot’ or ‘flash-bulb’ learning. One remembers them strongly after only a single instance. (Such memories may of course be strengthened by rehearsal – by telling the story over again, not always correctly.) Other types of memory benefit from repeated instances, from which one extracts the general nature of something, such as the meaning of an (undefined) word.
Procedural knowledge, such as driving a car, is often difficult to acquire from a single experience and usually benefits from repeated practice. It often lasts for a remarkably long time. Once you have learned to swim you can swim fairly well even if you haven’t swum for many years. A famous pianist said to me, about forgetting a familiar piece of music, ‘Muscle memory is the last to go,’ meaning by that term playing the piece automatically and without thinking about it.
Memories typically last for different times and are often divided into long-term and short-term memory, although the terms mean different things to different people. ‘Long-term’ usually means for hours, days, months, or even years. ‘Short-term’ can cover periods from a fraction of a second to a few minutes or more. Short-term memory is usually labile and of limited capacity.
Consider what happens when you are dreaming. It appears that you cannot put anything into long-term memory (or at least, anything you can explicitly recall) while you are actually dreaming. Your brain holds the dream in some form of short-term memory. When you wake up (which may happen more often than you realise) the long-term memory system switches on. Anything still in short-term memory can then be transferred into long term – that is why you remember, not everything you have dreamt, but the last few minutes of the dream. If, shortly after waking, you are disturbed – by a telephone call, for example – your short-term memory of the dream, being interrupted, may decay and be lost, so that after the telephone call is over you can no longer recall even the last part of your dream.
Recalling a memory, as we all know, is not a straightforward process. Usually some clue is needed to address the memory, and even then the memory may be elusive. Some memories become weak, and need stronger clues to evoke them. Others appear to fade until they are completely lost. A related memory may intrude and block access to the one you want, and so on.
From GERALD M. EDELMAN, ‘Building a Picture of the Brain’ (2001)
To say, as is commonplace, that memory involves storage raises the question: What is stored? Is it a coded message? When it is ‘read out’ or recovered, is it unchanged? These questions point to the widespread assumption that what is stored is some kind of representation. This in turn implies that the brain is supposed to be concerned with representations, at least in its cognitive functions. In perception, for example, even before memory occurs, alterations in the brain are supposed to stand for, symbolize, or portray what is experienced. In this view, memory is the more or less permanent laying down of changes that, when appropriately addressed, can recapture a representation – and, if necessary, act on it. In this view, learned acts are themselves the consequences of representations that store definite procedures or codes.
The idea that representational memory occurs in the brain carries with it a huge burden. While it allows an easy analogy to human informational transactions embedded in computers, that analogy poses more problems than it solves. In the case of humans working with computers, semantic operations occurring in the brain, not in the computer, are necessary to make sense of the coded syntactical strings that are stored physically in the computer either in a particular location or in a distributed form. Coherency must be maintained in the code (or error correction is required) and the capacity of the system is quite naturally expressed in terms of storage limits. Above all, the input to a computer must itself be coded in an unambiguous fashion; it must be syntactically ordered information.
The problem for the brain is that signals from the world do not in general represent a coded input. Instead, they are potentially ambiguous, are context-dependent, are subject to construction, and are not necessarily adorned by prior judgments as to their significance. An animal must categorize these signals for adaptive purposes, whether in perception or in memory, and somehow it must associate this categorization with subsequent experiences of the same kinds of signals. To do this with a coded or replicative storage system would require endless error correction, and a precision at least comparable to and possibly greater than that of computers. There is no evidence, however, that the structure of the brain could support such capabilities directly; neurons do not do floating-point arithmetic. It seems more likely that such mathematical capabilities have arisen in human culture as a consequence of symbolic exchange, linguistic interactions, and the application of logic.
Representation implies symbolic activity. This activity is at the center of our semantic and syntactical skills. It is no wonder that, in thinking about how the brain can repeat a performance, we are tempted to say that the brain represents. The flaws with such an assertion, however, are obvious: there is no precoded message in the signal, no structures capable of high-precision storage of a code, no judge in nature to provide decisions on alternative patterns, and no homunculus in the head to read a message. For these reasons, memory in the brain cannot be representational in the same way as it is in our devices.
What is it then, and how can one conceive of a nonrepresentational memory? In a complex brain, memory results from the selective matching that occurs between ongoing neural activity and signals from the world, the body, and the brain itself. The synaptic alterations that ensue affect the future responses of the brain to similar or different signals. These changes are reflected in the ability to repeat a mental or physical act in time and in a changing context. It is important here to indicate that by the word ‘act,’ I mean any ordered sequence of brain activities in a domain of perception, action, consciousness, speech, or even in the domain of meaning that in time leads to neural output. I stress time in my definition because it is the ability to recreate an act separated by a certain duration from the original signal set that is characteristic of memory. And in mentioning a changing context, I pay heed to a key property of memory in the brain: that it is, in some sense, a form of recategorization during ongoing experience rather than a precise replication of a previous sequence of events …
In this view, there are many hundreds, if not thousands, of separate memory systems in the brain. They range from all of the perceptual systems in different modalities to those systems governing intended or actual movement to those of the language system and speech sound. This gives recognition to the various types of memory tested by experimentalists in the field – procedural, semantic, episodic, and so on – but it does not restrict itself only to these types, which are defined mainly operationally and to some degree biochemically.
While individual memory systems differ, the key general conclusion is that memory itself is a system property. It cannot be equated solely to circuitry, synaptic changes, biochemistry, value constraints, or behavioral dynamics. Instead, it is the dynamic result of the interactions of all of these factors within a given system acting to select an output that repeats a performance. The overall characteristics of a particular performance may be similar to a previous performance within some threshold criterion, but the structures underlying any two similar performances can be quite different.
From JOHN MCCRONE, ‘Not-so Total Recall’, (2003)
Who are you? For most of us, our sense of self relies on a personal history of memories that can be dipped into just as readily as turning the pages of a photo album: the child who broke an arm falling out of a tree, the gawky teenager on a first date, the proud parent. But can your memory really be trusted with something as fundamental as your sense of identity?
Psychologists have long known that our memories are easily embellished. We add imaginary details through wishful thinking or to make a more logical story. More controversially, memory may be falsified through suggestion and manipulative questioning, bringing some eyewitness testimony and ‘recovered’ memories into doubt. And we all forget things too. But despite these flaws it was always presumed that the core experiences themselves – the memory traces stamped into the fabric of our brain – were permanent. Look in the right place and we could always dig back to what really happened.
