Skip to main content

How Are Memories Made?

Leonard Kelley holds a bachelor's in physics with a minor in mathematics. He loves the academic world and strives to constantly explore it.

This article will not be solving the mystery of memory, but it will present some findings that appeal to me. This can be because of their explanatory power but also for their original approaches to solving this enigma. I hope this gets you thinking more about those little moments that make up you, and what that really means.

Some Basics

Memory is broken into both short term and long term. With long term memory, we recall things from as recent as a few hours ago to years ago, while short term allows the “conscious manipulation of information.” Basically, you could make the analogy of long-term memory as data and short term as the data processor and not be too far off. We know from experience we can only recall a finite number of things that are presented to us at once because of the constant influx of stimuli, but is there a limit to how much we can store? Who knows, but maybe neuron networks can give us clues (Burnett 36-9).

Neurons send signals via action potentials across synapses, which are actually narrow gaps between neurons. Neurotransmitters from the pre-synaptic neuron traverse the synapses and cause electrical activity by activating receptor sites upon contacting the post-synaptic neuron. It is theorized that “synapses are the physical form for specific memories,” where the data is kept as it were. Our brain takes the information that has stuck into short term memory long enough and converts it into long term by encoding a new synapse (40-1).

Initial Inquires into Protein Memory

The road to getting to this basic understanding started in detailing brain scans. Once scientists started to use electroencephalograms (EEGs) to map brain activity, some began to ponder how the functional behavior lined up with the make up the brain. How did electrical impulses across synapses help contribute to the activity of the brain? (Rose 180-1)

Clearly, electrical activity was not correlated with long-term memory, but perhaps it functions at the short-term memory level. So how then does long term memory actually take place? Where does a single memory live? Can it be reduced? Can we mathematically determine the behavior? Is it located at discrete locations or over a broad area? Can cells processes the memories? (Ibid)

Most felt that microscopic measurements of structures would be key, allowing us to see the learning process in action. This has proven to be rather difficult to do, but scientists have been breaking down the scale into smaller and smaller pieces. A good starting point is the building blocks of the synapses themselves: Proteins and neurotransmitters. If we can see changes to these items, then perhaps we could gain overall structural changes and see ties to memory (181).

So let’s look at the most basic building block of the proteins and neurotransmitters, the molecule. Contrary to what intuition might tell you, a molecule inside of you actually stays fixed, instead undergoing some kind of change every few weeks or months. Cells and everything that is built upon them are constantly changing, and on the average half of all the protein molecules in you are revised every two weeks (182).

A protein inside of you is usually made over a few minutes and is sent to the part of the cell that needs it. Once at its destination, it will stay there anywhere from a few hours to a few months, then once it has been moved then it is removed from the cell via enzymes. They break down the protein into amino acids that will then be used for new proteins elsewhere in the body. For most people, the rate of protein production matches that of protein reduction (Ibid).

This led to the idea that if memory formation requires protein synthesis, then if we block that we should be able to block memories from being formed. By the early 1960s, protein behavior modification was discovered, and it involved those amino acids. Because a protein is a chain of these, we simply need to make one radioactive and we have an east way to track where it ends up and how far into the protein goes (183).

Therefore, the rate of growth of a protein should be correlated to the level of radioactivity seen in a given area. With 20 different amino acids out there and each protein being made of 100s of theme we certainly have lots of variability (Ibid).

The order of the amino acids is dependent on ribonucleic acid, or RNA, which is itself dependent on deoxynucleic acid, or DNA. So we now have another correlating factor to tie into protein production, giving us a large span of tools to try and see how these proteins are tied, if at all, to memory (Ibid).

One of the easiest ways to block protein production is through antibiotics, which halts RNA production. You could use this to stop production and see if someone can perform a task they recently learned. We could also use that radioactive tag, have someone perform a learning task, then compare the before and after and see the changes (184).

Enter the Hyden experiments, where rats were trained to balance on a wire in order to reach food. During the learning process, protein and RNA levels were up and it was noted where it happened in the brain. This allowed scientists to note where motor controls were. So scientists followed up by having rats try to eat food with their non-dominant hand by holding back the other one (Ibid).

This means they were doing a task they knew how to do but with different tools, so scientists knew what markers to look for and see the changes easily. And sure enough, the same changes in the dominant hand ended up happening in the non-dominant one, demonstrating protein/RNA correlations to learning (Ibid).

In the inhibitor experiments of Wesley Dingman and Michael Sporn, a water maze was used where rodents had to navigate obstacle sot get to a reward. An inhibitor of RNA synthesis was given at different points in the course. If it was given after the race, the rat did just fine. If it was given during the event, the rat failed the maze despite their training (Ibid).

Genetic Memory

This was encouraging, but the mechanics needed to be better parsed out, given a direction was it were. If we scale upward, we can view proteins, RNA, and DNA as macroscale quantities of these molecules that act as data processing. Could these be the information carriers of the brain? (189-191)

Extend this even further, thinking on an evolutionary scale where DNA carries the genetic information that is passed on from parent to offspring. If DNA really is involved in memory, could it be passed on genetically? The DNA/RNA/protein research seemed to point to this as a possibility, but how would one find evidence? (Ibid).

Scroll to Continue

James McConnel investigated this in the 1960s with the study of planaria, aka flatworms. He was able to train them to repsond to light by pairing that with electric shock. Once trained, they were killed, chopped up, and then fed to other flatworms which had no prior training. Talk about a nasty experiment! (191)

These new flatworms were then put through the same program, with no prior training, and had their responses measured. A set of flatworms that were untrained were also diced up and fed to new ones, which also went the rough program and acted as a control. The results were inconclusive, possibly due to the difficulty in training flatworms, but similar experiments with other animals seemed to point to promising results (Ibid).

In 1965, Allan Jacobson trained rats to go to a food dispenser when a flash of light or a click was made. These rats were killed, and their RNA was taken from the brain and placed into the gut cavity of an untrained rat. Once given the same stimuli, they followed the same behavior when prompted (191-2).

Ewen Cameron ran a study where elderly people with memory issues were given 100 grams of yeast RNA. Upon testing, it was found they had an increased performance in memory tasks, but was this because of the RNA or because they felt they were special for participating in the study and therefore felt it was more memorable? Also of concern was the lack of a control group nor an attempt at replication (192).

Georges Ungar had rodents released into an environment where it was lit everywhere except for a dark corner. If the rodent tried to go to the dark, it was given a shock and eventually it would stop choosing to go there. Material from their brains was given to new rodents and were allowed to make the same choice but with no shock treatment. Mice with the trained rodent material preferred to go to the light area and those with untrained material preferred the dark environment (193).

But this wasn’t enough for Ungar though, and he wanted to purify his samples and find the exact molecules which transferred the information. After several purification steps, he found that the active compound wasn’t RNA nor a protein but a peptide, which is 15-20 amino acids chained together (Ibid).

Naturally, this result received lots of criticism. How does this peptide carry the information? How would the peptide be sent to the right place for that specific memory to be transferred? Why would that peptide work across different rodent species? If it is really a component of memory allocation, then why haven’t more been seen in the brain? Was the response of the mice really learning or more of a flight-or-fight response to unknown stimuli? How does any of this actually resemble true genetic inheritance? Because of these concerns and a lack of replication, genetic memory research slowly died off (194-5).

Simple Nervous Systems

Researchers needed a way to actually witness and process how learning and memory were occurring but in a simple system. This is why aphasia were selected to see the cellular behavior of learning, and for years Eric Kandel has been the lead researcher with them (215).

Aplysia were selected for study because they have simple and limited behaviors, reducing the variable reactions possible. Their nervous system is easy to diagram and with only 20,000 neurons it makes mapping a much more achievable task. On top of that, these neurons are grouped very discretely, further allowing better data collection (Ibid).

For his study, Kandel looked at the gill and siphon withdrawal reflex isolated from the organism itself, claiming that how the neurons of this system interact and respond to artificial neurotransmitters should represent the memory for the reflex itself. He concluded based on the reflex itself, where touching the gill or siphon causes a retracting behavior (216-8).

About 50 sensory neurons are involved, connected to about 20 motor neurons directly for via intermediate neurons. Like many reactions, repetition yields diminished or eliminated responses unless other regions are activated, essentially rebooting our system. All of this operates on a short-term basis, of course (Ibid).

To better understand the process, Kandel needed to quantize the reaction. To do this, the slug was pinned down and a water pick used as a stimulus. Photocells were used to measure the retraction, and once the response was known it was needed to be seen how long until a habituated response was found. Then simply compare the activity via electricity. The mechanism is easy to extract, and the circuitry is relatively straight forward, removing potential convoluting factors such as peripheral neurons, circulatory neurotransmitters, and so forth (218-9).

By the mid-1970s, experiments showed that motor and sensory neurons were not where habituation was occurring because the expected decrease in electrical activity was not spotted at either locations. Somehow, the habituation was occurring in the interconnections between the sensory input and the motor neurons, somewhere in the synapses (219-220).

So somehow going from the pre to the post synapse is where the change is transmitted. Looking at the presynapse showed that 5-HT, aka serotoprin, was the neurotransmitter being sent. As habituation to a stimulus happens, increases, and decreases in how much the sensory presynapse terminal sends while the post synapse remains constant in its response (221).

Now, we do need to talk about some issues with trying to extend the results of Aplysia to larger, more complex systems. First, they are invertebrates, meaning their nervous system is vastly different from ours. Their neuron total is low, and their cells are larger while ours have lots of neurons, smaller cell sizes, and more interconnections between cells. Second, we removed so much from the host organism that we may not really be getting the full picture. Perhaps we accidentally are removing some critical interactions that we remain unaware of. We haven’t yet tested with a live aplasia yet either (224-6).

Where Is It Being Processed?

More complex organisms seem to operate differently. As far as what does the encoding or memory, that is likely the hippocampus, which is involved with sensory data. Tracking the movement of proteins in the region along with brain activity during sensory stimuli suggests that the encoding center is likely this structure (Burnett 41-3).

Upon being coded, the memory is then sent to the cortex, known as consolidation. So memories are stored in the brain, but how does the act of forgetting to occur then? Matter it’s a matter of use, a tie to an emotion that isn’t relevant, or an imperfection in the encoding or consolidation (Ibid).

Or we need to get more technical and break down the types of memories we have and see if that offers insights. After all, “no memory is formed in isolation” and so different situations trigger different memories. Episodic memory is tied to events, semantic memory is straight up information without a contextual basis, and procedural memory which are those auto-pilot skills done by us without consciously thinking about it (44-5).

Each of these has numerous examples of being important enough to merit repeated use and so make the transition from short term to long term. And of course each has its moments of use and so will not always be relevant in a current memory-recall. You may be trying to trace back to a prior moment with the memory but because of differences between then and now it renders recall difficulties (Ibid).

The Content Itself

In fact, this semantic versus episode memory hints at how difficult it is to truly convert memories from one person to another. Semantic memory is mainly datatory and so we can all reach a high-level of agreement over its contents but try to describe an episodic memory to someone. You can never fully convey the content of it because it required your unique brain’s structure, background, and emotional stance. Memories are more than a collection of atoms, instead about patterns and how they change their host structure. So, with different brains there is not way to fully communicate these special memories, making them very important to the individual (Yohan).

And then there are certain biases that we develop as we age that impact our memory’s reliability. So sometimes it’s not a matter of recall failure but of simply finding the original piece of data, which could have been updated as we grow and learn. This would then be a memory bias (Burnett 58-62).

Another encoding change can occur because you wish an event to be emotionally different as to support your self-esteem. This would be an egocentric bias. Ever remember needing to justify to yourself that you made a right choice? Well that would be a choice-supportive bias. The list is longer and more extensive than this, but I hope this gives you a general understanding of the potential pitfalls of memory (Ibid).

Repressed Memories

The creation of a memory can be a tricky thing, according to Richard McNally (Harvard psychologist), because it is an experience which is inherently subjective in its final recording. Those biases can come into play whether we realize it or not. The formation of the synapses is fixed into a state that has a strong emotional subtext encoded in it. This can lead to a memory being skewed and possibly falsified (Neimark 74).

This comes to the forefront when talking about a challenging topic like repressed memories. How can we test the accuracy of memories, especially when planting false ones has been confirmed? However, those are of trivial (i.e. not emotionally invested) events. If the event was emotionally charged, can you make a memory false whether it be by intent or by mistake? (Ibid)

McNally along with Susan Clancey and Daniel Schacter looked at a trauma group of women with reported resurfaced memories of sexual abuse. They compared these people to three other groups: those that had actually been abused and always remembered it, those that believed they had been abused but had no actual memory of it, and a control group of no abuse. All four groups were given an initial word-retrieval test and then another with another word added to the list (but without their knowledge) (74-5).

While I was unable to find data about two of those groups, the retrieved memory group reported the word was always there 68% of the time while the control group. While word choice and relevance to memory allocation could be objections, it is an interesting correlation. And when the study was done again but with alien abductees, the results paralleled the abuse one (Ibid)

As far as neural signatures go, one study by Stephen Kosslyn (Harvard University) looked at PET/fMRI scans of the brain and found that when people are asked to recall a visual memory, the vision pieces of the brain fire despite not actually seeing the object! And if the visual memory was especially emotional, the vision areas fired off even stronger. So emotionally charged events can lead to a better firing, but as far as accuracy goes that is still a debate (75).

Patient H.M.

Thus far, we have spoken somewhat broadly on memory, just to get our feet wet. But now it’s time to delve deeper. One of the most fascinating studies in memory dysfunction was Patient H.M. a 27-year-old who was first encountered by science in 1953 (Dowling 219).

This individual had “frequent and debilitating seizures” and the treatment at the time was taking out both hippocampi (for it was though the seizures arose from that part of the brain being diseased). Once this happened, Patient H.M.’s ability to develop long-term memories was eliminated and was therefore unable to retain any memories of events past the operation for long (Ibid).

Brenda Milner looked at H.M. for over 40 years and over that whole period the patient never remembered her each time they met. Yet this person was able to retain new motor abilities they developed despite being unable to recall learning said skills (Ibid).