memory

memory Life is unpredictable. But memory provides organisms with the ability to learn — to modify their behaviour in the light of experience — and hence to reduce their uncertainty about the world. This is clearly an important behavioural adaptation, from the point of view of evolution. Indeed, most animals exhibit some forms of learning and memory, ranging all the way from gradual weakening (habituation) or strengthening (sensitization) of simple reflex actions, to conscious recollection of personal experiences.

We can remember a telephone message for the few seconds it takes to write it down. But we can also remember things over very long periods of time. For example, adults may still remember some of the things they were taught at school — both general abilities, such as how to add numbers together, and specific things, such as the translation of ‘la plume de ma tante’. Additionally, we can also remember (though unconsciously) many of the skills attained through life, such as how to ride a bicycle or play the piano. There are many different ways in which humans and other animals remember things. It follows that memory cannot be conceptualized simply and that there are likely to be a variety of different, interacting memory systems.

Much of our sense of who we are as individuals depends on a particular kind of memory, involving recollection of our own past experiences, feelings, and relationships. One only has to imagine not being able to recall what has happened in one's past, or whether or not one even has family and friends, to realize how disruptive and distressing severe amnesia (such as occurs in Alzheimer's disease) can be, both to the patients themselves and to those close to them.

Psychologists have long drawn distinctions between different types of memory systems and memory processes. As early as 1890 William James distinguished between ‘primary memory’ (information one is presently aware of) and ‘secondary memory’ (information in the psychological past). Current ideas still maintain a distinction between short-term memory and long-term memory, evidenced by impairments of one or the other in brain-damaged patients. However, early ‘multistore models’, which proposed separate short-term and long-term memory stores, have now been discredited as being too simplistic.

The idea of a unitary short-term store has now largely been replaced by the concept of ‘working memory’. The working memory system is concerned with both active processing and short-term storage of information and allows one to plan for the future and to bring together thoughts and ideas. Damage to the frontal lobes seems to impair working memory: patients with such damage function rather normally apart from being impaired in the use of stored knowledge to guide appropriate behaviour. Experiments on monkeys have shown that individual nerve cells in certain parts of the frontal cortex not only fire impulses when certain objects are seen by the monkey but continue to respond when the object disappears from view, as if holding a memory of the object. Furthermore, studies on the effects of damage of the frontal lobes in monkeys suggest that different forms of working memory can be localized to specific regions of the prefrontal cortex — the front part of the frontal lobes.

The concept of a single long-term store has also been replaced, by the view that there are several interacting long-term memory systems. There have been many attempts to subdivide long-term memory, but none has proved entirely successful. Another early distinction was between ‘episodic memory’ and ‘semantic memory’. Episodic memory is autobiographical recollection of personally experienced events (such as what you had for breakfast), whereas semantic memory is general knowledge about the world, factual information and its meaning (such as the fact that breakfast is a kind of meal). Despite a clear conceptual difference, there is less evidence that these two types of memory rely on different memory systems in the brain. Indeed, semantic and episodic memory would appear to be strongly interdependent. For instance, retrieving semantic information may depend upon recalling the particular episodic event or events during which the semantic knowledge was gained. Likewise, it has been argued that recalling an episodic event (for example, remembering seeing an elephant at the zoo) depends on intact semantic memory (the definition of an elephant). Both types of memory may therefore rely on common underlying neural structures.

An alternative distinction was made between ‘declarative memory’ and ‘procedural memory’. Declarative memory refers to knowing ‘what’, and includes both semantic and episodic information, whereas procedural memory refers to knowing ‘how’, and relates to skilled behaviour without the need for conscious recollection, such as the ability to remember how to drive a car. Support comes from observation of certain patients with amnesia who seem to have relatively intact procedural learning abilities (they can still learn how to do things) in the face of impaired declarative learning (e.g. not remembering where they are). However, the distinction between declarative and procedural knowledge is imprecise and many kinds of behaviour involve aspects of both. Furthermore, some patients with severe amnesia are capable of certain feats of memory (such as learning new factual information) that cannot be explained by procedural learning alone.

A further theoretical distinction was made between ‘explicit memory’ and ‘implicit memory’. Explicit memory is said to be involved in tasks that require conscious recollection of previous experiences, whereas tasks that are facilitated in the absence of conscious recollection are said to depend on implicit memory. Many traditional methods used to test memory involve the person being asked to remember specific experiences, and are therefore measures of explicit memory. For instance, the memory of a previously-seen list of words could be tested by free recall (‘Tell me the words that were on that list you saw earlier’), by recognition (‘Was this word among the list you saw?’), or by cued recall (‘Complete these letters to form a word that occurred on the list’).

To demonstrate implicit memory it is necessary to show that a person has a long-term memory of a past experience although they can't consciously recall it. For example, the perceptual identification of words presented extremely briefly is easier if the words have previously been seen. Amnesic patients perform relatively normally on such ‘repetition priming’ tasks, as well as being able to acquire new motor skills, yet they are impaired on most tests of explicit memory. However, the distinction between explicit and implicit memory is again rather general and does not account for all of the patterns of long-term memory performance in amnesic subjects. Furthermore, the theory does nothing to address the fact that amnesic subjects can still form conscious short-term memories, which clearly involve explicit learning.

Observations that amnesic patients can retain some information briefly but not for long periods of time led to the development of the ‘consolidation theory’. This suggests that immediate experiences are somehow crystallized into long-term memory, and that this process is disrupted in amnesia. The theory also maintains that the process of memory consolidation occurs over a period of time, during which memory traces are particularly vulnerable to permanent disruption by such things as a blow to the head, certain drugs, electric shock to the brain, etc. However, consolidation theory cannot account for the fact that apparently lost memories can sometimes be retrieved subsequently.

‘Context-dependent theories’ on the other hand propose that each memory trace (for instance of a particular person) is encoded together with information about the associated context (where you met the person), and that subsequent retrieval of the memory may be facilitated by reinstating the context. (Everyone is familiar with the fact that it is difficult to remember the names of even close friends when you meet them in unexpected places.) This theory is supported by the remarkable observation that divers recall more words learnt underwater when subsequently tested underwater than when tested on land, and vice versa. Learning while under the influence of certain drugs is also context-dependent, being better recalled when the same drug is administered.

Related ‘state-dependent theories’ maintain that agents or procedures that induce amnesia do not permanently disrupt memories but rather ‘re-encode’ the memory traces in association with the brain state induced by the amnesic agent or procedure. Patients who have electroconvulsive shock (for instance, to treat depression) often complain of loss of memories; and this procedure indubitably disrupts long-term memory when given experimentally to rats. But rats can retrieve their lost memories after a subsequent shock, because this puts the brain back into the condition in which the information was ‘re-encoded’, thereby providing an additional cue to aid remembering.

Although it is hard to verify whether a deficiency of memory reflects re-encoding or permanent memory loss, the importance of forgetting should not be underestimated. Although the brain has a huge capacity for memories, it must be finite. Since the brain appears to be able to form associations between disparate stimuli very easily, so it is important for it to be able to forget meaningless or arbitrary associations and remember only those associations that prove consistent or relevant. It has been theorized that inappropriate associations in the brain may specifically be weakened during the phase of sleep in which rapid eye movements and vivid dreams occur (REM sleep).

It is intuitively obvious that memories of all sorts involve functional changes in the brain, sometimes occurring remarkably quickly. Much of what we know about learning and memory has been gained from clever experiments involving the training of animals, both intact and with brain damage, as well as from studies of normal and amnesic human beings. But over the past few decades neurophysiologists and molecular biologists have made great strides in their understanding of the cellular mechanisms of learning and memory. One fruitful approach has involved examining basic forms of learning in animals with relatively simple nervous systems, such as the marine snail Aplysia. This animal withdraws its gill apparatus reflexly when the ‘mantle’ around it is touched, and the circuit of sensory and motor nerve cells responsible for this has been defined. This reflex is subject to habituation (if the touch to the gill is repeated time after time), and to sensitization (if the touch is coupled with other stimulation).

It turns out that these simple forms of short-term implicit learning involve changes in the effectiveness of synaptic transmission (mainly changes in the amount of transmitter substance per nerve impulse released at a particular synapse in the circuit). Longer-term memory requires new protein synthesis and the growth of new or larger synapses.

More complicated forms of learning may involve elaboration of a common set of molecular mechanisms. For instance, most animals can learn to associate one stimulus with another (such as the association formed between the sound of a bell and the sight of food in Pavlovs' famous experiments on classical conditioning). The underlying neural change, just as for sensitization in Aplysia, is thought to involve increased release of transmitter substance at synapses in the circuit associating the two forms of stimulation.

In recent years, attention has focused on a primitive part of the cerebral cortex called the hippocampus, which is tucked inside, under the lower edge of the temporal lobe of the cerebral hemispheres. Extensive damage to this general region in humans can cause devastating retrograde amnesia, which virtually eliminates the capacity to form new long-term conscious memories, while leaving old semantic and personal memories relatively intact. Traditionally, the hippocampus itself has been considered the seat of human episodic memory. However, recent research with monkeys has revealed several, functionally dissociable memory systems in this region of the temporal lobe. These include the perirhinal cortex, for object memory, and the amygdala, for memory for the emotional significance of stimuli and events. These individual areas, each with its different specialization, may then contribute to a broader-based temporal lobe memory system providing the basis of both episodic and semantic memory. The monkey's hippocampus may have a relatively restricted role in memory for spatial location.

In rodents, the hippocampus certainly seems particularly involved in spatial memory: when it is damaged, rats and mice cannot remember their way around mazes. It turns out that the connections between certain nerve cells in the hippocampus are remarkably ‘plastic’. Synapses can be strengthened simply by a brief burst of nerve impulses, so that single impulses will subsequently (and for very long periods of time) evoke much bigger electrical responses in the receiving cell. Much is now known about the molecular basis of this phenomenon, called long-term potentiation. This mechanism may provide the basis of, or at least contribute to, many forms of learning, in several different regions of the brain, ranging from perceptual learning in young animals to human explicit memory.

Memory is central to the human condition and has been investigated at many levels. Neuroscientists have studied the molecular and cellular mechanisms of memory in animals and humans, and psychologists have contributed to our understanding about the different kinds of processes involved in memory through research with amnesic patients and normal subjects. Temporal lobe dysfunction is commonly associated with declarative or explicit memory impairments. However, since most amnesic patients either exhibit diffuse brain damage (Korsakoff's syndrome) or have focal damage to a range of different structures, our present understanding of which particular neural systems are important for different memory processes has come predominantly from animal ‘models’ of human amnesia.

Mark J. Buckley

Bibliography

Bolhuis, J. (2000). Brain, perception, memory: advances in cognitive neuroscience. Oxford University Press.
Eysenck, M. W. (1995). Cognitive psychology: a student's handbook, (3rd edn). Erlbaum, Hove.

See also amnesia; brain; cerebral cortex; limbic system.