2.17 Memory

Two-year-old Sophie had never seen snow before. Together with her mother Amy, she was attending her first science discussion group, in deep midwinter. Amy had re-joined the group after a break and was curious about her child’s developing memory. “Sophie clearly knows whether or not she had seen something before; she must have a short-term memory” she inferred. “She knows what’s coming, she anticipates things. When do babies develop short term and long-term memory?” she wondered.

Julie, always keen to dig a bit deeper, wondered what memory actually is, for any of us, at any age: “how is a memory held in the brain” she asked.  “It must be complicated because people recall the same thing in different ways. Some remember smells, others, sights or sounds. Amy remembers conversations, her sister remembers what things look like. Julie remembers whether she likes a book but forgets the plot. “I think I’d read that you tend to remember what is most important to you” Marian chipped in.


Working out what happens when we remember things is a key theme in experimental psychology. Cleverly designed experiments with humans and other animals have established a number of key concepts. The most important, alluded to by Amy, is that there are both long and short-term storage systems. Tests on the short-term component suggest that it retains material for only a few tens of seconds, sufficiently long to remember a phone number, for example. According to the “working memory” model, information held temporarily in this short-term area interacts with information already held in the long-term memory. This is inferred from experiments that show, for example, that it’s easier to remember a sequence of eight digits if they correspond to two memorable years for you (e.g. 19141918) rather than a random sequence. This link to long-term memory is vital as the storage capacity of the short-term memory is strictly limited: somewhere between five and seven items (plus or minus two) at a time. A single item can, of course, consist of many smaller items; for example, it’s much easier to remember five six-letter words than thirty random letters.

Long term memory clearly has the capacity to store many disparate kinds of thing, from people and places to skills and emotions. Two broad classifications have been developed by psychologists. So-called “procedural” memory includes the things we retrieve automatically, without recourse to conscious control. Musicians, gymnasts and typists are obvious examples of people who have committed complex physical actions to memory through repetitive training; but so too have the rest of us, when we learn to tie our shoelaces or rustle up an omelette.

A different kind of long-term memory, known as “declarative” memory, includes facts and events that require conscious recall. There are various sub-categories of declarative memory. For example, “episodic” memory enables us to recall sensations and emotions from a particular place or time in our lives that relate to episodes we have experienced personally. A different type of memory called “semantic” plays a big part in school learning and pub quizzes. It’s associated with abstract factual knowledge, like scores in the last football World Cup or Latin names of garden plants.

The distinct types of memory capacity in the brain have been identified historically through experiments in psychology and studies of various kinds of impairment or injury. Patrick mentioned a concert pianist he had known who had suffered from early-onset Alzheimer’s disease. Even when no longer able to recognise her own daughter, she would confidently play through a Beethoven piano sonata from memory. Clearly different parts of the brain anatomy are responsible for different kinds of memory, such as visual, auditory and language memory, for example. Neuroscience research attempts to relate differences in memory function to particular parts of the brain.

One component of the brain called the hippocampus is found to be involved in declarative memory and in consolidating memories (i.e. transferring from short-term to long-term memory). It’s also known to be associated with spatial awareness. A well-publicised study of trained London taxi drivers showed part of their hippocampus to be enlarged as a consequence of their knowledge of the street plan. A related study showed the opposite effect for blind people. 

Buried deep inside the brain, the two hippocampi act as a kind of hub, picking up signals from many parts of the outer brain (known as the cortex) – such as the visual cortex – and sending out signals to various other parts.

Figure 1 Location of the hippocampus

A different part, known as the amygdala, also deep within the organ, is known to be associated with emotional memory. For example, when a state of fear is aroused by your senses perceiving a threat, associations are formed in the two amygdalae with memories of this threat.

Figure 2. Location of the amygdalae

Our understanding of the role of the hippocampus was accelerated by the case of a US patient in the 1950s who suffered with severe epilepsy. His hippocampus was removed to halt his frequent seizures. Subsequently, he was unable to form new memories, despite being able to recall events prior to the operation. His working memory (short-term) and procedural memory (skills), however, remained intact. It’s evidence of this kind, coupled with findings from experimental studies, that help us separate out each distinct aspect of memory and identify the brain systems that support them.

Misremembering and forgetting

Sarah chipped in at this point with a further complication. “It’s surely possible to misremember things. You can make mistakes recollecting things” she stated, to nods of agreement. “Witnesses to crimes are sometimes found to have had false memories” she added. Julie added to the complexity by mentioning trauma, in which things get buried but must still be stored somewhere in your memory even though not accessible. “It’s about access as much as storage”, she concluded.

Research in psychology offers a number of competing theories about this – it’s not a settled matter. What seems to be generally agreed is that interpretation plays a role in recalling a memory. A memory may be inaccessible because it has been repressed. Equally it may be modified as it’s retrieved, by extraneous material from the context of the original incident. Two things may get confused, so that, for example, a witness may recall a person’s presence at a particular scene when they had, in fact, been present on a different occasion. Memory recall is clearly not a simple process of retrieval, as it is in a computer.

“What about forgetting” interjected Julie at this point in the exploration. “My mum had been a potter all her life, but as dementia overtook her, she completely forgot how to pot”. Again, psychology offers more than one theory about this process. One idea, that forgetting may be the result of memory simply decaying over time, remains unresolved, as it’s so hard to design studies to test it. An alternative idea, better supported by evidence, is that items already in short-term memory are simply pushed out by newer ones. There is only room for a few items at a time in this space; the oldest one has to go, as a new one enters.

Longer term memory involves retrieval as well as storage. A key idea is that cues are required to retrieve a memory, and these can simply be missing when we fail to recall something. Having something on the tip of the tongue is indicative of this – a sense that you know you know something but can’t quite trigger the memory. Cues often involve a reminder of the context in which the memory was laid down – like a smell from long ago or a visit to your old school. Another idea is that the retrieval process can be interfered with by a subsequent action related to the original memory. An example is forgetting how to use manual gears in a car, after you’ve got used to an automatic gearbox.

There’s also the simple fact that memories are stored, not in some abstract space, but in the biological material of our brains. Neuroscientists are looking at ways in which the structures and chemical processes within our brains give rise to forgetting, as well as remembering. One current idea is that forgetting may not be a failing at all, but an important active process in the brain, clearing the way for new impressions. Imagine how cluttered your mind would become if every detail of your bus journey were recalled – the clothes every passenger was wearing, the expressions on their faces, the comings and goings at each stop.


Julie’s fundamental question still irked Patrick: “how are memories actually stored in the brain”. Results from psychology experiments may tell us how people (and other animals) respond in given situations, but they aren’t able to explain exactly how the stuff of the brain actually handles cognitive actions like thinking and remembering. The more recent discipline of neuroscience is beginning to address some of these, thanks to developments in technology. Powerful microscopes are strong enough to show individual brain cells, enabling researchers to physically manipulate them, to some extent. MRI machines create real-time images of the points at which blood is flowing in the brain, indicating those parts that are active when a subject performs a specific task.

Storage and retrieval of information has become an everyday issue today, thanks to the ubiquity of digital devices. We take for granted their ability to store and retrieve information with (almost) perfect fidelity. The fundamental units of which they are made – transistors – can only exist in one of two states: “on” or “off”.  Labelling these two states 1 and 0 enables numbers to be stored in binary form. Letters, lines, colours and shapes can be represented using these binary numbers as codes – the code for the letter “a”, for example, is the binary number 01100001. Images and text can thus be stored digitally, then retrieved perfectly and brought to life on a digital screen when required. Computers don’t forget or misremember!

The brain works in utterly different ways to a computer, as anyone who has failed to recapture a memory will know! In broad terms, however, the similarities are worth noting. The fundamental units of which the brain is composed are, like transistors on a microchip, extremely small. They are also present in their billions. These units are the brain cells, of which there are several types. The main ones involved in memory are nerve cells or neurons, which have the special ability to transmit electrical pulses.

As shown in the microscope photograph in figure 3, a neuron has a central body (which keeps the cell alive and active) from which a plethora of filaments extend. These are the channels – known as dendrites – through which electrical pulses from other cells arrive at the cell body.

Figure 3 A single neuron (from a rat brain)

A single neuron may have hundreds or even thousands of these dendrites.  In addition, one thicker tentacle extending from the cell body, known as the axon, enables the cell to send out electrical pulses to neighbouring cells.

Through the dendrites, one cell may be receiving pulses from hundreds or thousands of other cells. This is just one indication of the extraordinary connectivity of the network of neurons in the brain (figure 4). Equally astounding is the sheer number of neurons in a human brain – in the region of 100 billion.

Figure 4 network of connected neurons (Axons are red, dendrites green)

Combining this huge number with the number of connections each cell makes, gives some idea of how much more complex the brain is than even the most sophisticated computer, with a mere 1 or 2 billion transistors.

It’s not just the much lower storage capacity that distinguishes a computer from a brain; it’s also its mode of operation. A computer follows, blindly and precisely, a set of instructions written by human beings, using data supplied by human intervention. The brain does not slavishly follow a programme, nor is controlled by an external agent. It functions dynamically by continuously developing networks of neurons. It achieves this by strengthening or weakening the multiple connections between them. One neuron will pass on an electrical pulse to a neighbouring cell, through one of its many dendrite extensions, if, and only if, the strength of the pulse it receives exceeds a given threshold. With input pulses arriving at a cell from many such dendrites, that cell may or may not reach a threshold required to fire off an output pulse. Thus, several factors determine whether a given neuron in a network gets stimulated by its neighbours to fire. The complexity of the way in which pulses, running through groups of neurons, give rise to particular memories and feelings, is not fully understood. Some important aspects of the process are however becoming known. A key one is the role of repetition.


The output arm of a neuron – the axon – carries an electrical pulse from the body of the cell to the tip – the most distant part of the cell. To participate in a network, it then needs to connect with a neighbouring cell.

The place where the output of one cell (tip of an axon) meets the input of a neighbour (a dendrite) is a junction known as a synapse (figure 5). The pulse is not carried over electrically, however, as it is when two metal wires are connected in a circuit. Instead, a flotilla of chemicals is released by the first neuron which traverses the gap between and is taken up by the second one.

Figure 5 Diagram of a synapse

This then sends on the pulse electrically through the second cell. Because they transmit the signal from one neuron to the next, these chemicals are known as neurotransmitters. Many kinds of neurotransmitter exist; some of which are becoming well known though their association with brain disorders: serotonin, histamine, dopamine and adrenalin, for example.

A key discovery in recent years has been that transmission at synapses gets strengthened or weakened, according to use. If the same network is activated repeatedly, its synapses (junctions) grow stronger. It is this process that explains the effectiveness of repetitive practice in sport, music, typing, driving or any activity in which we learn by doing. It underlies the way in which, as babies, infants and children we acquire skills and knowledge. By repeatedly falling over, manipulating objects and listening to language, we gradually learn to associate external experiences with inner representations in neural networks. It is also, as you might expect, how we gradually lose competences, as the connections between neurons weaken or cease functioning with age.

Memories are somehow represented in networks of neurons, whose interconnections have been made strong. It is in the process of transferring memories of experience from the short-term to the long-term parts of the system that this strengthening occurs. It’s not easy to see how this biological level of explanation can account for the extraordinary variety of remembered experience, from the aesthetics of a glorious sunset to the horror of a traumatic moment. But this model does at least give us a sense of how memories can be held more or less strongly and how some may be stored so firmly and located so specifically that they can be recalled even with dementia. Mention of this extraordinary aspect of brain disease prompted Sarah to tell the group about her aged mother who, as she succumbed to Alzheimer’s disease, recalled the German language she had learned as a child, despite having emigrated from Germany at the age of eight.

As an aside, the design of today’s Artificial Intelligence software is inspired by networks of neurons in the brain. So-called “artificial neural networks” consist of computer memory units linked together in vast networks with multiple connections between each unit. The thresholds (or weightings) which determine whether one unit or another will be switched on by the other units to which it is connected are determined through repeated cycles of trial and error, mimicking the behaviour of biological neurons. This approach is proving helpful in many practical situations, including, for example, screening medical scans. An AI network can be trained to work out the threshold values (or weightings) that correspond to a positive identification of breast cancer, for example, by presenting it with a large number of breast cancer X ray images. Thereafter it will use these values to spot further cases without human intervention.

Brain conditions associated with difficulties in remembering, reading, maths and music are explored below in Further Reading: Dementia, dyslexia and other conditions

Back to babies

This exploration began with a young mother reflecting on the memory capacity of her baby. Research backs up her observation that babies have a short-term memory, more or less from the beginning. We know that a baby recognises their mother’s voice at birth and, soon after, the taste of her breast milk. But, as we also know from experience, they will quickly forget what they have learned. Eight-week-old babies forget within a few hours, according to one study, while three-month olds can remember for up to a week. As they grow, we also see infants gradually learning, and committing to memory, important information about the size and shape and physical properties of objects around them. It is not until they are 3 or 4 years old, however, that their memory begins to record things that have happened and who was present at the time. Although, as adults we are not able to remember experiences from our earliest years, young children can. For example, a six-year-old is capable of remembering events from before their first birthday, but that capacity gets lost with age.

In an ingenious experiment with mirrors and a dab of red on the nose, researchers found that infants began to recognise themselves around the age of two. The children were able to link the red nose in the mirror with the red they knew was on their nose. At an earlier age they may have imagined the person in the mirror to be someone else. Self-recognition at his age is the starting point for social awareness and the subsequent capacity to develop relationships.


Our ability to remember and forget is a constant topic of conversation. Simple curiosity may be the trigger, or fascination with a growing child or worry about an ageing relative. In the world of criminal investigation or PTSD treatment, it assumes major importance. Our understanding of it depends on experiments on people and animals by psychologists and laboratory studies of brain material by neuroscientists. Combining these two areas of study, we now understand much about the processes of remembering and forgetting. It nevertheless remains utterly remarkable to me, and largely unexplained, how, in our hasty lives, the aroma of baking bread and snatches of a distant accordion manage to evoke happy memories of a long-past holiday!

© Andrew Morris 2023

Further Reading – Dementia, dyslexia and other conditions


The image we now have of the brain as a vast network of interconnected cells helps us understand how things can go wrong. Clearly messages will not be passed on within a network if neurons die off or fail to operate. This is what happens in the case of Alzheimer’s disease.

Proteins that are part of the normal structure of a cell get broken down into fragments in the space between cells; normally, these are washed away. In the Alzheimer’s condition they remain present and clump together to form plaques (brown in figure 6) between neurons, disrupting the activity of the cells.

Figure 6 Illustration of neurons in Alzheimer’s disease (Amyloid plaques in brown; tangles in blue).

In addition, a different protein in the cell fails to form properly and accumulates in tangles (blue) within neurons. These two problems impair the functioning of the neurons causing, amongst other things, loss of memory.

Older members of the discussion group were keen to point out that remembering things seems to get increasingly difficult with age anyway, regardless of disease. Studies show this to be true, with declining ability in both short-term working memory and recall of longer-term stored-up memories. This is an active area of research with suggestions that it is linked to loss of function in the hippocampus area, known to be important for memory. Gradual loss of cells with age may be the cause or decreasing levels of the neurotransmitters that connect one cell to another. Other studies show that, whereas younger animals are able to recover full cognitive functioning after being subjected to stress, older ones, are less able to. Brain-cell structures become more rigid with age. These factors give us some indication of why cognitive capacity declines with age and learning is impaired.

Dyslexia, dyscalculia and amusia

Discussion of impairments in the brain inspired Sarah to tell the group about her experience of dyslexia. She had been on a course about the condition and was surprised to learn what a high percentage of the population have dyslexia. Dyslexia UK and the NHS estimate the incidence in the UK at around 10% but higher estimates are also reported around the world.  A recent UK report finds that schools are failing to report some 80% of cases. Sarah had learned that the difficulty for people with dyslexia lies with the phonemes – the smallest units of sound in language –  which need to be remembered accurately. Languages where the sounds correspond closely to the written letters, such as Spanish, Italian and Finnish, are easier for people with dyslexia to learn. Research using fMRI scans shows there can be reduced activity in the language processing and phonological (sound processing) areas of the brain in people with dyslexia. Other studies, comparing twins brought up in different environments, suggest there is also a genetic component to the condition. 

Patrick, who had once been a maths teacher, picked up on an analogous condition – dyscalculia –  for people who had difficulty processing mathematical concepts. He had taught many otherwise capable students who had major problems grasping concepts of number and symbolism in algebra. There appears to be no clear understanding of the biological causes of dyscalculia, but there are studies on how children develop mathematic abilities in general.

Babies of three months are able to differentiate between large and small quantities. At school we learn words to associate with quantities. From this we develop the concept of a “number line” which enables us to estimate quantities and place them in rank order. Initially, as we struggle with the novelty of these concepts, our abilities develop in the thinking areas of the brain, where we hold short-term memories. With time and practice, however, they are gradually transferred to longer-term memory areas, where their use becomes automatic. This latter process happens much more slowly for children with dyscalculia. They therefore require much greater concentration than their peers to process simple mathematical operations; they may for instance have to count-up a quantity of things one by one, where others are able to simply grasp it in a glance.

At this point the discussion group began speculating about other cognitive processes that could be impaired by brain functioning – music for instance. Some people describe themselves as ‘tone deaf’ or lacking musical ability. It turns out there is indeed a condition known as ‘amusia’, a disorder which a person cannot judge or reproduce the pitch of a musical sound. It appears that babies are usually born with a musical sense, being responsive to scales and rhythm.  Infants sense the difference between harmonious and discordant sounds. It seems that training is not necessary to grasp the harmony embodied in chords or the sequence of notes in a scale or to respond to differences in key. These develop naturally just with exposure to music – but not for the 4% with amusia. For them, it may not be possible to detect that a wrong note has been played. From studies of people with various kinds of brain damage, it is clear that several distinct networks exist in the brain for different aspects of musical processing. Some can grasp the interval between a high and lower pitch note but not discern whether the music is tonal or not. Others may be able to remember spoken words but not melodies or the other way round. A sense of pulse – keeping to a beat – may be difficult for some and memorising a short segment of rhythm, impossible for others.

© Andrew Morris 2023