🕒 17 min
Learning is the process of acquiring new knowledge, skills, and behaviour. It follows us throughout our lives and enables us to adapt to various new situations.
Memory is the ability of storing the knowledge, skills, and behaviour acquired through learning in our brains. We can differentiate between explicit and implicit memory, as well as sensoric, short-term, and long-term memory. These types of memory are frequently misunderstood, so I will try to explain them as easily as possible.
1. Types of memory
Explicit memory is a type of memory that requires active memorization or extra work, for example, studying for an exam. It is dependant on the hippocampus, a special part of the brain whose role in memory will be explained in later sections. Explicit memory can be further divided into two different types – episodic and semantic.
Episodic memory is for remembering different events in our lives – birthday parties, your first day of school, yesterday’s lunch – the various episodes of our lives.
Semantic memory is for remembering abstract facts. Semantics is just a complicated word for “meaning”.
Implicit memory is a type of memory that does not need active remembering patterns. It is also mostly known for procedural memory, the memorization of movements and motor skills such as riding a bike or roller-skating. There are other types of implicit memory (such as priming and conditioning), but they are not as important to us currently. Implicit memory is highly dependant on the basal ganglia and the cerebellum, special parts of our brain which regulate our movement.
We can also differentiate memory into types by how long it stays stored in our brains. The shortest-lived one is called sensory memory because it is directly connected to the types of sensory stimuli we experience – sounds, smells, touch. It lasts as long as we are feeling the stimulus; as soon as that goes away, we cannot remember all the aspects of it.
Our short-term memory lasts a bit longer than sensory memory. It can store information as long as it is being constantly repeated, like random passwords or phone numbers which we just read.
Then there is long-term memory, the kind of memory which can last for years and even decades. It depends on the process of consolidation, the process of making memories, which will be explained later.
2. A short introduction to our brain cells
A long time ago, in the 19th century, neuroscience was still in its early stages, especially regarding our understanding of the structure of the central nervous system (CNS). There were two main competing conceptions: neuronal doctrine and reticular theory.
Neuronal doctrine postulated that the CNS is made of many small cells called neurons which are in close proximity to one another, but do not share their intracellular contents, also known as the cytoplasm.
On the other hand, reticular theory claimed that neurons make up what is called a syncytium – a heap of physically connected cells which share their cytoplasm.
You may be wondering how the scientists of that era could have had such different thoughts about the subject at hand. Couldn’t they have just looked at brain slices under the microscope to find out what neurons actually looked like?
Well, they could have, but they would not have found anything useful. Brain tissue is so densely packed with various cells, not just neurons, that in that mess of cells you cannot really tell if they are physically separated or not. That is why they needed to come up with a method of looking at the neurons themselves, and only a handful of them.
One of the loudest and most famous reticularists was Camillo Golgi, an Italian neuroscientist who came up with a special reaction that could dye only 1% of the neurons in black. He called it la reazione nera (black reaction) and he got some pretty marvellous pictures of the hippocampus which he used to prove his theory. In his name, the black reaction is also known as the Golgi stain.
At the same time, there was also Santiago Ramón y Cajal, a Spanish neuroscientist who was a very big opponent of the reticular theory. His extensive research of the cerebellum yielded some of the most beautiful Golgi stain pictures of the Purkinje cells (cells that make sure that our movements are smooth and precise). Cajal was a very smart man and quite far ahead of his time, so he used the Golgi method to prove Golgi wrong – that neurons are separate entities, and not part of one big neuron.
His main proof was the fact that dying neurons did not cause the death of other neurons around them if a toxin was injected. If the reticular theory were correct, the toxin would pass on from one neuron to the next, as they share a cytoplasm. Since that was not the case, this was one of many proofs to come to disprove the reticular theory. Unfortunately, it would take almost 200 years until the first electron microscope definitively showed that neurons do not form a syncytium.
2.1. Neuron anatomy
So, since we now know that neurons exist, we can ask ourselves: what do they look like?
Neurons consist of the same typical cellular organelles that a lot of other cells have, but the morphology (how they actually look) is vastly different because of their function – to receive and send chemical and electrical signals.
Dendrites are little projections through which neurons receive inputs from other neurons. They can be of various lengths and widths, often with many bifurcations. Some dendrites we call “mossy” because they have moss-like spines on them. These spines are the places where the neuron accepts input from other cells. The processes of learning and memorization are established by changing the morphology of the spines, among various other processes.
The cell body is the place where all the inputs are summed up and where most of the protein and neurotransmitter synthesis happens.
The axonal hillock is where every neuron asks itself a very Hamletian question, “To send a signal or not to send a signal?”
The axon is a cellular part whose function is to send the fired signal from the body of the neuron to its end, where it will likely serve as an input for the dendrites of another neuron. Contrary to popular belief, an axon does not have to be one long strand, it can have many strands (also known as collaterals), but they all come off the main axon.
2.2. What is a synapse?
The place where a neuron passes a signal on to another neuron is called a synapse. It is the space between the axonal ending of the presynaptic neuron and a dendritic spine or part of a dendrite, cell body or even an axon of the postsynaptic neuron. A typical synapse works as follows:
- An impulse arrives to the end of an axon.
- The axonal ending sends a neurotransmitter into the synapse.
- The neurotransmitter connects onto a receptor on the dendrite.
- The receptor causes an influx (or efflux) of ions which causes a change in the intracellular charge.
- That change in the charge is summed with other changes on other dendrites and if the sum is sufficiently high, the signal is sent from the cell through the axon onto other cells.
3. Learning mechanisms
3.1. Synaptic plasticity – how our synapses change
We can see that synapses are dynamic structures and that they consist of several moving parts. What can we change in a synapse to change it properties?
- We could increase or decrease the amount of neurotransmitter ejected by the presynaptic neuron.
- We could increase or decrease the number of receptors on the postsynaptic neuron.
By playing with these two mechanisms, we can very precisely tune how strong the synapse will be and how likely it is to set off a signal in other cells. One of the most famous ways that the strength of a synapse changes is known as long-term potentiation or LTP.
3.2. Long-term potentiation (LTP)
By using the LTP mechanism, a synapse can increase in strength and retain that effect for hours and even days.
To understand what LTP is, we need to dive a bit deeper into biochemistry. We will not go too far, but we need to know who the main players are in this game. They are: one neurotransmitter, two ions, and two receptors.
Glutamate is a neurotransmitter. It is ejected by the presynaptic neuron when the signal comes to the end of an axon.
Sodium (Na+) and calcium (Ca2+) ions are the ones that have a crucial role in LTP.
AMPA receptor (“AMPA” is an abbreviation of α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid, so I think it is best if we stick with AMPA) is a receptor which opens and lets Na+ ions into the cell when glutamate binds to it. NMDA receptor (“NMDA” is short for N-methyl-D-aspartate, which is not as bad as AMPA, but still too long for neuroscientists, so we will just call them NMDA), on the other hand, opens and lets Ca2+ ions into the cell in response to glutamate.
Now that we know all of this, we can look at the mechanism of LTP. It is divided into two stages: the early stage and the late stage.
Early stage LTP
- Glutamate exits the presynaptic neuron and binds to the AMPA and NMDA receptors.
- NMDA receptor allows Ca2+ ions to enter the dendritic spine and AMPA receptors allow Na+ ions in.
- Ca2+ ions increase the charge inside the cell which opens other receptors and lets a lot of Na+ ions into the cell.
- Na+ ions increase the charge and can set off a signal to other neurons.
Why is this important? Because we needed to get Ca2+ into the cell! Calcium is the one which activates all the enzymes that induce the start of the LTP process. What do those enzymes do?
- They increase the number of AMPA receptors, which increases the influx of Na+.
- They change the AMPA receptors so that they stay open longer, which additionally furthers Na+ influx.
Late stage LTP
The late stage manifests as an activation of a bigger network of enzymes which causes the synthesis of new receptors. It can also change the shape and size of the dendritic spine and create retrograde signals by sending various chemicals from the postsynaptic neuron into the presynaptic neuron and change the amount of neurotransmitter it ejects.
4. Memory mechanisms
By knowing how one synapse can change, now we can change multiple synapses in a neural network and make them do something, like remembering facts.
Generally speaking, to remember stuff we need to learn it first. Some things we can learn easily, while others we need to repeat a couple of times before they are stored in our memory. Both of those go through a process called consolidation.
Consolidation is a process by which temporary (labile) memories are transformed into a more permanent (stable) form also known as an engram (memory trace).
When we experience something, the complex memory system in our brain combines a vast amount of sensory and cognitive inputs – sight, smell, taste, where we are, the position of our body, the people around us, and a lot of other stuff – into one cohesive event which has some context attributed to it (when, where and how it happened) so we could remember it more easily. And, if you remember the types of memory from the beginning of this text, that type of memory is called episodic memory.
This model is also known as the standard model of consolidation. The process of consolidation would then be the process of strengthening the bonds in the cortex of the brain between the cells which store our memories. These cells are located in the hippocampus, a structure inside the brain dedicated to the process of consolidation.
The hippocampus is an important structure which creates episodic memory and transforms short-term into long-term memory. How does it do that? Actually quite elegantly!
Neuroscientists uncovered that there are engram cells in the hippocampus – neurons which activate during an event and then change their physical and chemical properties (synaptic plasticity!) and as such “entangle” themselves with the event which changed them. This means they can reactivate later when the event repeats. By reactivating the engram cell, the memory of the event is more likely to be remembered (like when we are trying to memorize something). It can also be deactivated when we forget something (for example, a traumatic experience or a fact which we have not consolidated).
Experiments done on mice, primates and people who suffer from amnesia have shown that amnesia more frequently hits memories which are have been established recently, rather than those which were formed a long time ago. This makes sense if we remember that the more recent the memory is, the less likely it is to have gone through the process of consolidation.
A theory suggests that the hippocampus creates an “index” or a template of neural activity which was present at the time of the event we are trying to remember, but the information of the event is stored elsewhere. Remembering involves getting the information about the event from the hippocampus, which is dedicated to the process of remembering the event, into the cortex, which is dedicated to storing the information of the event.
Multiple trace theory tells us that the hippocampus is always used when we try to recollect something, but that the existence of multiple engram cells which point to the same cell, which stores the event, is why we can recollect some events better than others.
This is how we can explain how we remember abstract ideas and why we can make semantic memory. Firstly, we remember an abstract idea in the context of one event. Then another event creates an engram cell which point to the same cell which stores the same abstract idea. By having multiple cells which all point to the same storage cell, we can remember the pattern of activation and thus remember the abstract idea without the hippocampus.
Why? Because, as was mentioned earlier, the hippocampus gives context to memories. For something abstract we do not need context, just the idea – so we do not need the hippocampus. For example, to recollect some episodic memory (like our 10th birthday party) we need the context in which we can remember it – our emotions during the day, which of our friends were there, what kind of cake we had etc.
4.3. Engrams and neural networks
Now you know that, after a while, we can remember episodic memories without the need for the hippocampus. How is that possible when the hippocampus is the very structure that made the memory?
That is because the memory exists in another place in the brain which can act and activate independently of the hippocampus. This place is called an engram. The name was formulated in 1904 by Richard Simon when he postulated that an engram is:
- a chemical or physical change in a network of neurons
- which occurs because of the activation of other neurons
- and can be reactivated by activating the same neurons (which causes remembrance)
So how does this fit into our consolidation theory? Easily:
- The hippocampus receives all sensory and cognitive inputs and forms a memory.
- That memory is temporarily stored in the hippocampus and surrounding structures, where it is only accessible as short-term memory.
- By the process of consolidation, the memory is moved from the hippocampus into the prefrontal cortex, where it stays until the memory is forgotten.
If we think back to the process of LTP, now it should be clear what its is purpose here. It strengthens the bonds between the neurons in charge of remembering a particular event.
If we wanted to localize where and how exactly the memories are stored in our brains, we would need to learn about three important structures:
- Origin: the hippocampus (the medial entorhinal cortex or MEC)
- Destination 1: the medial prefrontal cortex or mPFC
- Destination 2: the basolateral amygdala or BLA
When the memory is first formed in the hippocampus as sensory or short-term memory, the engram cell in the hippocampus sends an impulse to the mPFC and “reserves” an engram cell there for this new memory. These “reserved” engram cells are also known as “silent” engram cells because they still do not posses any information. Simultaneously, an impulse is sent to the BLA engram cells which store the emotional context of the memory. This is very nicely symbolically represented in the animated movie Inside Out, where different memories have different colours depending on what kind of emotional memory they are – happy, sad, angry, scary or disgusting.
How do silent engram cells differ from activated engram cells? Well, silent engram cells have a small number of dendritic spines, which means they have fewer synapses and lower activity. That is why we cannot remember the stuff we learn completely the day after we learnt it.
4.4. Engram cell maturation
Repetition is one of the most important methods we can use to ensure that we remember something. This is because, through the process of repetition, consolidation takes place and the number of signals coming into the silent engram cells increases, causing the formation of new spines and stronger synapses (LTP again!). This is how a silent engram cell becomes an active engram cell.
Recently, it was discovered that the connection between silent and active engram cells is a two-way one – silent engram cells could even transform active engram cells into silent ones!
The maturation from silent to active engram cells is fueled by the constant influx of information from the hippocampus. Once the consolidation process is done and the engram cell in the mPFC is mature, there is no need for the existence of the hippocampal engram cell, so it goes back to its original silent state. The mechanism of this is unknown – probably the two-way connection between the mPFC and hippocampal engram cells.
If we look at the mPFC and hippocampal cells two weeks after learning, we can see that recalling a memory takes place in the prefrontal engram cells, and not the hippocampal ones. That is, unless one has been reminded of the learned fact – then the hippocampal engram cell stays active.
One of the proposed solutions to the enigma of disappearing hippocampal engram cells lies in the fact that the hippocampus is where new neurons are made (one of very few such places) and that those new neurons are the cause of amnesia or forgetting because they disturb the delicate structure of hippocampal neural networks.
By now you should be screaming at me because you know we can easily remember highly emotional events (deaths, break-ups, first kisses) without any need of consolidation. How is that possible?
Because that is the main role of the BLA engram cells. They are special in the sense that they never return to their silent mode. Once the BLA engram cell is established and entangled with a specific emotional event, it maintains a very strong bond with its corresponding prefrontal engram cell.
5. Future research
I hope I have made the processes of learning and memory as clear as I could have through a blog post like this. Researchers are looking into a lot that I was not able to cover here, but which you might be interested in pondering over (since, as far as I’m aware, there are no research papers on these subjects):
- When do hippocampal engram cells turn silent prefrontal engram cells into active ones – the role of sleep in the process of memorization?
- How do hippocampal engram cells return to their silent state – recurrent connections between prefrontal and hippocampal cells?
- Are episodic and semantic memory both stored in mPFC engram cells or are they separate?
If you have any ideas about these, feel free to comment and I would be very glad to see what you came up with!
Additional reading and sources
- A detailed seminar in which I explain the same subject and answer questions from the audience (Croatian only)
- LTP explained through animation
- Consolidation explained (older video, does not have all the newest information)
- Original research paper on engram cells
- Pictures of the brain are pretty, but nothing beats the live (or dead) brain