Where is my mind? A brief review of neuroscientists’ search for biological substrates of memory

“Life is all memory, except for the one present moment that goes by you so quickly you hardly catch it going.” - Tennessee Williams 

The capacity to remember is not only undeniably vital for our survival, but the unique stream of memories we each save also make up the essence of who we are as individuals. We can remember to never again touch a hot stove after getting burnt and effortlessly recall where and how to procure food, so as not to starve. Memory also gives us the ability to remember the names and faces of our friends and foes, and allows us to relive our most cherished and haunting moments in vivid detail. This remarkable capability of “mental time-travel” is likely taken for granted on a daily basis, but has been a topic of heavy scientific investigation for over a century. In fact, mankind’s interest in the subject dates at least as far back as 350 B.C. when Aristotle described memory as “the scribe of the soul”. The human brain is estimated to be comprised of approximately 100 billion neurons [1], each of which potentially form synaptic connections with up to 10,000 other neurons, arriving us somewhere in the ballpark of a quadrillion (or 10^15) synapses. In other words, the number of synapses in your brain far exceeds the number of observable stars in the Milky Way galaxy. Amidst all of this commotion, our brains need to appropriately organize themselves to allow us to do things like move, speak, perceive the world around us and somehow save all of this information for future use. So how does it work?

In the late 1800s, seminal studies done by Santiago Ramón y Cajal, now known as the father of modern neuroscience, challenged the widely held belief that the brain was made up of one continuous reticular substance. He instead argued that the brain was in fact made up of many discrete cells, or neurons, that communicated to each other across small gaps between them (although at this time it was unknown to him how this communication was accomplished, and it was in fact the neurophysiologist Charles Scott Sherrington, not Cajal, who ascribed the term “synapse” to these gaps in 1897). It wasn’t until the early 1900s that Cajal’s ideas began to firmly take hold, ultimately earning Cajal the Nobel Prize in 1911. Subsequently, scientists began to theorize that experience could alter the efficacy of cellular communication across these synapses, and thereby store information within them. A few decades after Cajal’s paradigm-shifting studies, Donald Hebb postulated that neurons that are activated in close temporal proximity to one another will tend to have stronger (or potentiated) synaptic connections than those that are activated far apart in time [2]. This phenomenon, now known as Hebbian plasticity (colloquially described by the rhyme “cells that fire together wire together”) sparked research aimed at understanding how synaptic strengthening occurs at a molecular and cellular level, and whether it might truly have some role in the formation and retention of memories. A landmark study by Timothy Bliss & Terje Lomo over two decades later definitively showed long-term potentiation (LTP) of synapses in the hippocampus of the mammalian brain [3], and a great many studies demonstrating the importance of LTP in learning and memory followed suit. To many in the field, these experiments suggested that the mechanisms of synaptic potentiation must be the cellular basis of memory itself. To this day, many scientists still regard LTP as the most attractive candidate for the cellular substrate of memory, and firmly support the idea that memories are stored solely by altering synaptic strength within brain circuits.

Around the same time that the Hebbian postulate came to light, Karl Lashley was performing a number of experiments on rodents of the following type: systematically destroy a section of the cortex and study what effect losing that brain region has on the animal’s ability to perform memory-related tasks.  Remarkably, he found in almost all cases that the animals were able to perform just fine despite having a portion of its brain obliterated, and noticed that it only seemed to matter how much cortex was destroyed rather than specifically what portion. From his studies, summarized in his influential 1950 publication [4], Lashley developed a firm opposition to the idea that memories were localized, and instead suggested that they must be distributed broadly throughout the cortex, bringing new life to the notion of a “memory engram” or mnemic trace, conceptualized years earlier by the evolutionary biologist Richard Semon. Specifically, the idea behind the memory engram was that in response to environmental stimuli, some biochemical or biophysical change occurs in many cells across the cortex which allows for the storage of memory. Importantly, the engram does not exist solely in isolated connections between input and output cells, according to Lashley, thus adding another layer of complexity to the emerging insights on synaptic storage of memory. Scientific technology has improved dramatically since the 1950’s, which has allowed memory researchers to move away from “destroy-and-see-what-happens” methods toward more elegant approaches, in turn leading to strong evidence in support of the engram hypothesis. A testament to the advancement of neuroscientific technology can be seen in the work of modern researchers like Tomás Ryan, who have used clever techniques to show that experimenter-induced amnesia could be overcome by artificially stimulating engram cells in rodents, effectively recovering lost memories [5].

The pieces of the puzzle may appear to be falling nicely into place – by merging the two concepts presented thus far, we can imagine how a group of cells distributed across the brain might be active during an event-to-be-remembered (i.e. engram cells) which, through experiencing the event-to-be-remembered, strengthen their connections with one another via LTP or similar mechanisms. Voila! We have figured out how memory works, right? Well, almost. While experiments on engram cells and synaptic strengthening have undoubtedly advanced our knowledge of how memory works, there are still some stubborn puzzle pieces that don’t quite fit…

For example, a major question mark at this point is whether synapse identity is a critical component of memory. Put differently, do individual synapses support distinct memories? And by extension, if a synapse is lost, does the memory it supports also get lost?  While the jury is still out on this issue, we do know that cells and synapses are intrinsically dynamic -- synapses routinely form and retract throughout life, and there is constant turnover of the proteins and architectural elements that compose synapses, requiring replacements to be made and transported to the correct sites. This replacement process relies on transcribing genes and turning those gene transcripts into functional proteins that are necessary for the formation and maintenance of synapses (more on this later). The question then becomes: if a particular synapse that is strengthened during the formation of a memory needs to be maintained in order for the memory to survive, how is it possible to keep that synapse alive and strong if the proteins that compose it are constantly degrading and needing to be replaced? A real-life analogy to help think of how this process might work is as follows: Let’s say you have an electronic device that runs on a battery. After using the device for some time, the battery dies, causing you to look for and install a replacement. Perhaps with repeated use of the device, other parts also begin to fail, so you venture out and find replacements for those as well. In this example, the constituent parts of the device may have changed (i.e. there is a new battery in place) but the function of the device remains the same once the new pieces have been installed. You could presumably continue using the device and replacing the worn out parts in such a way for your entire life. Similarly, neurons could continue using and replacing the constituent parts of synapses for your entire life, all while maintaining the function of storing memory.

While using and replacing components of specific synapses in order to maintain a memory appears to be feasible, it doesn’t necessarily answer our question of whether synapse identity is critically important. Furthermore, it begs the question of whether evolution would select for this process, which at face value seems to be fairly labile, to support a crucially important function like memory. As I mentioned, cells need to transcribe genes and make proteins in order to perform the cellular analog of our battery replacement example. When neurons become strongly activated, signals from the synapse are sent to the nucleus of the cell where DNA is stored as chromatin (a fancy word for tightly packed-up DNA). These molecular signals, which can come in the form of proteins, enzymes and even non-coding RNAs, can then tell the chromatin to express more of the genes that are important for maintaining strong synapses. However, chromatin is so tightly packed into the microscopic space of the nucleus that in order to access the regions of DNA that are needed to make your cellular “replacement batteries”, the cell needs to employ epigenetic mechanisms.

In short, epigenetic mechanisms cause the chromatin to become more relaxed or condensed, which allows genes of interest to be more highly expressed or repressed, respectively. What is borne out of this line of research is the possibility that memory, in its simplest form, may be operationalized as specific changes to the chromatin landscape within the nucleus of cells that serve as a code for which genes and proteins are expressed by that cell, which in turn will depend on its prior activity and connections with other neurons. Furthermore, depending on the strength and/or frequency of a particular experience, the extent of chromatin modifications and the relative length of time that they last might differ (which would then correlate to the strength and persistence of different memories). In other words, there may be stable, experience-dependent “engrams” embedded into the very genome of each neuron in your brain that ultimately allow your memories to be maintained over the course of your lifetime. Importantly, this view does not rule out the importance of changing synaptic strength during memory formation, but instead posits that the physical substrate of the memory engram lies within the nucleus of the cell, rather than at the synapses themselves.

An intriguing study from David Glanzman’s lab earlier in 2018 provides some support for this notion [6]. Alexis Bédécarrats and colleagues showed that they could transfer sensitization memory from a trained sea-snail to an untrained one, simply by injecting RNA from the trained snail into the naïve animal. The interpretation was that RNA from the trained animal was able to induce an epigenetic engram in the recipient animal, allowing it to express the memory of the donor snail. (A brief aside: sensitization and an information-rich memory of tasting delicious gelato on the streets of Rome, for example, are by no means the same thing. However, the reader should note that they are both bona fide forms of memory and are supported by the same cellular mechanisms). Still other work has shown that hibernating animals have significant loss of synapses as a result of decreased body temperature during hibernation, which rapidly recovers once the animal comes out of hibernation [7]. This is noteworthy because these animals do not exit hibernation with amnesia, despite the significant loss and regeneration of synapses. If memories were explicitly stored in specific synapses, neither of the above phenomena would be possible.  

It goes without saying that much more work needs to be done to comprehensively understand how memory works, but it is likely the case that memory is supported by some amalgamation of the concepts presented here, perhaps unified by some heretofore undiscovered process. We know that synapses between cells that are coincidentally activated become strengthened in order to form and encode memories, thereby creating the memory engram. Concerted epigenetic changes then likely result across the engram cells, allowing memories to remain stable over time, even in the face of dynamic cellular change and synaptic turnover. But many important questions still loom without answers, like what is the precise molecular form of the signals that travel from synapse to nucleus, and nucleus to synapse? Beyond that, how do those molecules exert their epigenetic actions to alter the way the genome is expressed in different cells? What determines whether those changes are transient or permanent? And for the ambitious, through our developing understanding of memory and its biological bases, might there be a way to restore memory to those who have lost it? Though there is still a long way to go and much work to be done to answer these and other questions, it is undoubtedly fascinating to reflect on where the decades of memory research has gotten us, and incredibly exciting to think about where it may take us in the future.


1)      Herculano-Houzel, S. (2009). The human brain in numbers: a linearly scaled-up primate brain. Frontiers in Human Neuroscience, 3, 31.

2)      Hebb, D. O. The organization of behavior. New York: Wiley; (1949)

3)      Bliss, T. V. P., & Lømo, T. (1973). Long-lasting potentiation of synaptic transmission in the dentate area of the anaesthetized rabbit following stimulation of the perforant path. The Journal of Physiology232(2), 331–356.

4)      Lashley, K. S. (1950). In search of the engram. In Society for Experimental Biology, Physiological mechanisms in animal behavior. (Society's Symposium IV.) (pp. 454-482). Oxford, England: Academic Press.

5)      Ryan, T. J., Roy, D. S., Pignatelli, M., Arons, A., & Tonegawa, S. (2015). Memory. Engram cells retain memory under retrograde amnesia. Science (New York, N.Y.), 348(6238), 1007–13.

6)      Bédécarrats, A., Chen, S., Pearce, K., Cai, D., & Glanzman, D. L. (2018). RNA from Trained Aplysia Can Induce an Epigenetic Engram for Long-Term Sensitization in Untrained Aplysia. Eneuro, 5(3), ENEURO.0038-18.2018.

7)      von der Ohe, C. G., Garner, C. C., Darian-Smith, C., & Heller, H. C. (2007). Synaptic protein dynamics in hibernation. The Journal of Neuroscience : The Official Journal of the Society for Neuroscience, 27(1), 84–92.