Knowable Magazine · Memory, the mystery

Memory, the mystery: Just in the past half-century, our understanding of how exactly our brains remember has taken huge leaps. Amazingly, this is just the beginning.


Listen on: Amazon Music | Apple Podcasts | Breaker | Deezer | Google Podcasts | LibSyn | Spotify | Stitcher

Introducing the Knowable podcast. What are the limits to what’s knowable — and how does our thinking about big questions in science and technology evolve over time? Take an audio journey with Knowable Magazine from Annual Reviews as we explore puzzles as diverse as the existence of black holes and how to build an artificial heart — with plenty of surprises along the way. More episodes


How can science connect something intangible and abstract to something physical? What can these connections teach us about our experiences of the world? And what does it mean when these physical mechanisms are faulty?

This is Knowable, and I’m Adam Levy.

How far back can you remember? For me, I can picture myself sat in the corridor of my grandparents’ house, dwarfed by the passing grown-ups. I remember my brother doing a silly dance to coax me out of a temper tantrum. And, of course, I remember that one kid who wouldn’t share the tricycle with me at kindergarten. But where actually are these memories? I don’t mean, where did the events take place? I mean, how are they actually stored? And how far can I trust my storage system?

Well, we don’t have to cast our minds very far back before the answer to these questions was a resounding “nobody knows”. Even as late as the 1970s, it was quite unclear what memory physically was. We know that memory exists on a personal level, of course, because we, well, we remember things. And we can see that other animals remember things too, since past events influence the future behavior of creatures from worms to orangutans:

This change in behavior indicates that learning has occurred and has left a memory trace somewhere, presumably in the brain but not necessarily.

Not necessarily in the brain! This is from a 1972 review titled “Effects of Chemical and Physical Treatments on Learning and Memory” and published in the Annual Review of Psychology. The review also clarified why memory is such a fundamentally difficult problem. Memory has to be studied — somehow — in action, or in laying down a memory to begin with. In the living brain of an organism, rather than as cells in a test tube or petri dish. And studying the brain — if that was indeed where memory was stored — is hard, since …

… cerebral processes un­derlying learning and memory are very well insulated from the outside world.

Around the middle of the 20th century, ideas were still hazy about what memory was made of, and where and how it was stored.

Sheena Josselyn: “Back then, in the 1950s and ’60s, ideas were sort of all that we had. We didn’t have the tools and the techniques that we have today to try to answer some of these questions. There were a lot, a lot of questions, a lot of theories, and not too much testing of the theories.”

This is neuroscientist Sheena Josselyn, who’s based at the Hospital for Sick Children in Toronto, and the University of Toronto. So what were some of these theories? Well, one concerned the material in which memory is stored. A 1967 review titled simply “Memory” and published in the Annual Review of Medicine stated that it was conceivable …

… that memory is specifically encoded in macromolecules, in some fashion anal­ogous to genetic encoding. 

This is completely different from our current understanding, and just half a century ago. Today researchers understand that memory is somehow written in the vast number of connections — synapses — between neurons in the brain. These connections can strengthen through use, a process that is widely believed to encode the memory.

But 50 years ago there was scant evidence of these processes. And so it wasn’t much of a stretch to suppose that our memories could be written in coded molecules, in much the same way as genes are written in our DNA. Memories could be stored in RNA, for example. What’s more, there seemed to be experimental justification for this view.

The 1967 review we heard from a minute ago details experiments where flatworms were fed the remains of their deceased companions, and seemed to learn something from their ill-fated flatworm friends. While these results still tantalize researchers today, in the 1960s this was being taken one step further:

Danish investigators have reported successful transfer from intracisternal injections of RNA from rats trained to approach either light or dark for a water reward.

Yes, that’s right. Researchers believed they had successfully injected memories of a task — in the form of RNA — from one rat to another. Taking RNA from a rat that was trained to approach dark or light, and injecting it into an untrained animal, seemed to transfer the skill from one rat to the other. Labs even attempted to inject or transfer the memory of a maze from one rat to another, taking these experiments to the next level.

Larry Squire: “It seems preposterous in a way now, but at the time it was studied pretty extensively and there were positive claims.”

Neuroscientist Larry Squire of the University of California San Diego.

Another key question was not over what memory was made of, but precisely where it was stored. For decades, in the early 20th century, psychologist Karl Lashley hunted for the trace of memories in the brains of trained rats. He lesioned — in other words, damaged — their brains in specific places, investigating whether he could cut a specific memory out. But whatever he did, he couldn’t get rid of, and so locate, the memory trace, also called an engram.

Sheena Josselyn: “So he was one of the first people who said ‘OK, engrams are there, but they’re nowhere.’ He could never find one specific spot.”

This led to the belief that memories are not stored in one location, but instead — somehow — stored everywhere across the brain.

Larry Squire: “The general idea was that there’s no particular region of the brain thought to be important for memory specifically. But it all changed in the middle of the 20th century.”

A major driving force behind this change was one individual. Not a scientist — a patient. American Henry Molaison, referred to in the literature simply as H.M. In 1953, H.M. underwent surgery in which parts of his brain were removed in the hopes of relieving his epileptic seizures. The operation was a success. To an extent. But the surgery came with a serious side effect.

Larry Squire: “He became profoundly, profoundly memory-impaired, but in the absence of any other significant change in his perceptual abilities, his intellectual abilities, his personality. So that really inaugurated, one can say, the modern era of memory research by showing there’s a particular part of the brain important for memory.”

Through H.M. — as well as evidence from other patients — it became clear that memory’s operation took place somewhere, not dispersed throughout the entire brain. It’s now thought that Karl Lashley couldn’t find the precise location of memories because he was searching in the wrong places, or because of the complexity of the memories he was hoping to erase.

But the patient H.M. had more to reveal than that memories were formed using certain key regions of the brain. It appeared that while H.M. was unable to consciously remember new facts, he was still able to learn. In 1982, Larry — who we just heard from — penned a review titled “The Neuropsychology of Human Memory” for the Annual Review of Neuroscience, which noted:

The brain regions damaged in amnesia, while necessary for many or most kinds of learning and memory, may not be required for certain other kinds.

H.M. was able to learn and improve certain motor skills — such as hand-eye coordination tasks — even if he had no memory of having practiced them before. Along with evidence from other patients, this indicated new — albeit unconscious — information was still being stored in the brain, and stored completely separately to conscious memories of facts and experiences.


Larry Squire: “Yeah, to my mind, that was one of the major advances in the last quarter century, because they pointed out that memory is not a singular faculty of the mind, there are multiple forms of memory.”

While patients were teaching researchers that memory had different forms, and the importance of different brain regions, other researchers began to untangle the physical form of memory. Some scientists suggested that memory traces — or engrams — could take on a very different shape from the molecules that had previously been proposed to store them.

By 1981, the idea had begun to take shape. A review in the Annual Review of Pharmacology and Toxicology (also by Larry), titled “The Pharmacology of Memory: A Neurobiological Perspective,” explained:

The development of long-term memory appears to occur over a period of time and might involve biochemical events that alter synaptic connectivity in a stable way.

While the review pointed out that these physical mechanisms had been linked with certain behaviors …

… none of these has yet been clearly linked to behavioral memory.

Gradually, though, this idea strengthened. A decade later, in 1993, a review titled “Structural Changes Accompanying Memory Storage” for the Annual Review of Physiology stated that:

Perhaps the most striking finding in the biology of memory is that long-term memory involves structural changes.

And these structural changes consisted of …

… alterations in the structure of synaptic connections.

Today, this has become the widely accepted explanation for how memory is stored. But this is far from the end of the story for understanding memory’s physical form:

Sheena Josselyn: “We think that it’s somehow within the circuit of cells that memories are held. We’re still, you know, very vague on the details, we hand-wave and things like that.”

But researchers are beginning to tease out further details of the mind’s memory storage. This research goes beyond the “where,” and begins to probe the “how.”

Larry Squire: “Eventually one wants to work out not just which structures are the important ones, but also how exactly they operate and what computations they actually carry out. In order to do that, one can come in with these newer techniques.”

These new techniques include such innovations as functional magnetic resonance imaging (fMRI) and positron emission tomography (PET). While studies of the damaged brain — either in animals or in human subjects like H.M. — provide deep insights, these imaging tools allow for …

… the visualization of memory processes in the healthy brain. Functional neuroimaging studies allow for the design of psychological experiments targeted at specific memory processes …

… according to the 1998 review “Cognitive Neuroscience of Human Memory” in the Annual Review of Psychology. These tools have been incredibly useful in identifying the location and operation of memory on larger scales. But they measure blood flow or metabolism in the brain, rather than neural activity itself, and …

… the indirect measure of neural activity limits the temporal and spatial fidelity of activations.

In other words, while imaging has provided insights into memories, these approaches can’t tell us much about what’s happening on the level of individual neurons and their connections. But in recent years, one particular technology has arisen that has allowed researchers — like Sheena who we heard from earlier — to study the brain’s function with unprecedented detail.

Sheena Josselyn: “One of the really cool ones that we use is called optogenetics, where we can turn a cell basically on or off just by shining different wavelengths of light into the brain of a rodent, which is just sort of game-changing. So now we can control individual neurons while a rodent is behaving.”

Optogenetics allow researchers to program cells so that they can control them remotely while the organism is otherwise functioning normally. Techniques like these allow researchers to begin to study how engrams — those memory traces — take shape and function within the brain. That includes, for example, beginning to understand how neurons are recruited to become a part of an engram. And how the process can go, well, wrong.

In 2018, Sheena co-authored a review on this topic for the Annual Review of Neuroscience, where she discusses not just the formation of engrams but also misinformation errors. Memory mistakes, she and her coauthor wrote, might arise from the processes that lead to some neurons being included — and others excluded — from the physical memory storage.

Such ideas begin to get at one of the most tantalizing psychological revelations of the 20th century: that memory can be distorted and even outright fabricated. Memory, it turns out, does not work like a cassette tape through which we simply rewind and replay past events.

Sheena Josselyn: “My daughter probably doesn’t even know what a cassette tape is but, yes, memories are not like cassette tapes. They are not a permanent, you know, video recording of what actually happened. They can be very constructive, which means that sometimes I can remember a memory slightly differently than it happened.”

Eyewitness testimony, for example, is riddled with false memories — a known problem in the legal system.

Anyone who’s ever had a blazing row about who said they’d wash the dishes knows that memory is fallible. And now, physical investigations of the brain can begin to tell us not only how memories are laid down and are reinforced. But also how they change, and how they are forgotten.

Larry Squire: “Is forgetting actually involving the physical loss of the substrate of the memory itself, as if, like in the melting of an ice cube or the loss of branches from a tree? Is that the situation, or is it the case that it’s all based on interference and that everything is really there if one could find methods of bringing it back?”

We now know by looking at the physical changes in the brains of simpler animals that memories truly can be lost — no longer stored in the connections of our brains. What’s more, our understanding of memory has advanced to such a level that experiments are now able to insert memories. No, not through injections of RNA, but through careful manipulation of neuronal connections.

Sheena Josselyn: “So, some amazing work from Paul Frankland’s lab — and full disclosure, he’s my husband — where he implanted an entirely false memory inside the brain of a mouse. So this event never actually happened, but the mouse behaved as if it actually had, just because he could use optogenetics and he knew a little bit about engrams and how a memory would be formed, he could implant an entirely false memory in the brain of a mouse. It’s sort of science fiction.”

Experiments like these shine a light on just how far we’ve come since the middle of the 20th century.

Larry Squire: “One only has to look back 10 years, 20 years, 30 years, when we had a completely different conception of things. And before 1950, we really knew nothing about it and, in fact, a lot of things we thought we knew were wrong.”

But even though research has come a long way, there’s still a long way to go. And as researchers add more tools to their belt, they’ll get an ever better handle on how memory operates in the brain.

Larry Squire: “As new techniques become available, we can ask questions that we couldn’t even ask before. At the very least, we’ll look back and say, ‘We really didn’t know much back then.’”

For Sheena, this progress is invaluable, not least because it may help us one day apply our understanding to the treatment of patients.

Sheena Josselyn: “I think the real challenge is to take the really cool things that we can do in the lab and try and make the most out of this knowledge and apply it to the human condition. My real quest is to understand how humans learn to remember things, so that people who have disorders of memory, we can start to help them out.”

This could mean learning how to dampen memories for people suffering from post-traumatic stress disorder, or help reinforce memories for those with dementia.

Such treatments may require even deeper theoretical understanding of how memory operates. After all, even with all the tools and achievements of the last half century, we are still a long way from connecting the experience of a memory to what takes place in our brains.

Sheena Josselyn: “What do we still have left to learn about memory? And the answer is, ‘A lot.’ I think we still don’t know where a memory lies. Is it in sort of emerging brain activity? Is it in how these different cells that make up an engram fire? How is it that, you know, past behavior can guide future behavior in something like, you know, a flatworm up to something like a human? The final textbook on memory has yet to be written.”

That’s the end of Season 1 of the Knowable podcast, but there’s more to come. Make sure you’re subscribed on your favorite podcast app so you don’t miss Season 2, which we hope to have out later this year. Don’t miss it!

In the meantime, be sure to drop us an unforgettable review wherever you get your pods, and send us your feedback to [email protected].

This podcast was made by Knowable Magazine from Annual Reviews, a journalistic endeavor dedicated to making scientific knowledge accessible to all. Through smart storytelling and sound science, Knowable aims to build understanding and fascination with the world around us. Knowable Magazine is free, and always will be. Read more at 

In this episode you heard from Sheena Josselyn and Larry Squire. There were also quotes from six papers: Murray E. Jarvik, 1972; Ward C. Halstead et al., 1967; Larry R. Squire et al., 1981 and 1982; Craig H. Bailey and Eric R. Kandel, 1993; and J.D.E. Gabrieli, 1998. I’m Adam Levy, and this has been Knowable.