Neuroengineers has found the first evidence that individual neurons in the human brain target specific memories during recall


An important aspect of human memory is our ability to conjure specific moments from the vast array of experiences that have occurred in any given setting.

For example, if asked to recommend a tourist itinerary for a city you have visited many times, your brain somehow enables you to selectively recall and distinguish specific memories from your different trips to provide an answer.

Studies have shown that declarative memory – the kind of memory you can consciously recall like your home address or your mother’s name – relies on healthy medial temporal lobe structures in the brain, including the hippocampus and entorhinal cortex (EC).

These regions are also important for spatial cognition, demonstrated by the Nobel-Prize-winning discovery of “place cells” and “grid cells” in these regions—neurons that activate to represent specific locations in the environment during navigation (akin to a GPS).

However, it has not been clear if or how this “spatial map” in the brain relates to a person’s memory of events at those locations, and how neuronal activity in these regions enables us to target a particular memory for retrieval among related experiences.

A team led by neuroengineers at Columbia Engineering has found the first evidence that individual neurons in the human brain target specific memories during recall.

They studied recordings in neurosurgical patients who had electrodes implanted in their brains and examined how the patients’ brain signals corresponded to their behavior while performing a virtual-reality (VR) object-location memory task.

The researchers identified “memory-trace cells” whose activity was spatially tuned to the location where subjects remembered encountering specific objects. The study is published today in Nature Neuroscience.

Video of example trials of spatial memory task showing memory encoding and retrieval. Credit: Salman Qasim/Columbia Engineering

“We found these memory-trace neurons primarily in the entorhinal cortex (EC), which is one of the first regions of the brain affected by the onset of Alzheimer ‘s disease,” says Joshua Jacobs, associate professor of biomedical engineering, who directed the study.

“Because the activity of these neurons is closely related to what a person is trying to remember, it is possible that their activity is disrupted by diseases like Alzheimer’s, leading to memory deficits.

Our findings should open up new lines of investigation into how neural activity in the entorhinal cortex and medial temporal lobe helps us target past events for recall, and more generally how space and memory overlap in the brain.”

The team was able to measure the activity of single neurons by taking advantage of a rare opportunity: invasively recording from the brains of 19 neurosurgical patients at several hospitals, including the Columbia University Irving Medical Center.

The patients had drug-resistant epilepsy and so had already had recording electrodes implanted in their brains for their clinical treatment.

The researchers designed experiments as engaging and immersive VR computer games and the bedridden patients used laptops and handheld controllers to move through virtual environments.

In performing the task, subjects first navigated through the environment to learn the locations of four unique objects.

Then the researchers removed the objects and asked patients to move through the environment and mark the location of one specific object on each trial.

The team measured the activity of neurons as the patients moved through the environment and marked their memory targets.

Initially, they identified purely spatially tuned neurons similar to “place cells” that always activated when patients moved through specific locations, regardless of the subjects’ memory target.

“These neurons seemed only to care about the person’s spatial location, like a pure GPS,” says Salman E. Qasim, Jacobs’ Ph.D. student and lead author of the study.

Specific neurons that map memories now identified in the human brain
Conceptual illustration of neurons that “map” memories in the human brain. Credit: Salman Qasim/Columbia Engineering

But the researchers also noticed that other neurons only activated in locations relevant to the memory the patient was recalling on that trial—whenever patients were instructed to target a different memory for recall, these neurons changed their activity to match the new target’s remembered location.

What especially excited Jacobs and Qasim is that they could actually decode the specific memory a patient was targeting based on the activity of these neurons.

“Our study demonstrates that neurons in the human brain track the experiences we are willfully recalling, and can change their activity patterns to differentiate between memories. They’re just like the pins on your Google map that mark the locations you remember for important events,” Qasim says.

“This discovery might provide a potential mechanism for our ability to selectively call upon different experiences from the past and highlights how these memories may influence our brain’s spatial map.”

Jacobs and Qasim plan next to look for evidence that these neurons represent memories in non-spatial contexts to better characterize their role in memory function.

“We know now that neurons care about where our memories occur and now we want to see if these neurons care about other features of those memories, like when they occurred, what occurred, and so on,” Qasim notes.

Short-term memory (STM), also referred to as short-term storage, or primary or active memory indicates different systems of memory involved in the retention of pieces of information (memory chunks) for a relatively short time (usually up to 30 seconds). In contrast, long-term memory (LTM) may hold an indefinite amount of information. The difference between the two memories, however, is not just in the ‘time’ variable but is above all functional.

Nevertheless, the two systems are closely related. Practically, STM works as a kind of “scratchpad” for temporary recall of a limited number of data (in the verbal domain, roughly the George Miller’s ‘magical’ number 7 +/- 2 items) that come from the sensory register and are ready to be processed through attention and recognition.[1] 

On the other side, information collected in the LTM storage consist of memories for the performance of actions or skills (i.e., procedural memories, “knowing how”) and memories of facts, rules, concepts, and events (i.e., declarative memories, “knowing that”).

Declarative memory includes semantic and episodic memory. The former concerns broad knowledge of facts, rules, concepts, and propositions (‘general knowledge’), the latter is related to personal and experienced events and the contexts in which they occurred (‘personal recollection’).

Although STM is closely related to the concept of ‘working memory’ (WM), STM and WM represent two distinct entities. STM, indeed, is a set of storage systems whereas WM indicates the cognitive operations and executive functions associated with the organization and manipulation of stored information.[2] Nevertheless, one hears the terms STM and WM often used interchangeably.

Furthermore, one must distinguish STM from the ‘sensory memory’ (SM) such as the acoustical echoic and iconic visual memories which are shorter in duration (fraction of a second) than STM and reflect the original sensation, or perception, of the stimulus. In other words, SM is specific to the stimulus’ modality of presentation. This ‘raw’ sensory information undergoes processing, and when it becomes STM gets expressed in a format different from that perceived initially.

The famous Atkinson and Shiffrin model (or multi-store model), proposed in the late 1960s, explains the functional correlations between STM, LTM, SM, and WM.[3] Later on, a considerable number of studies demonstrated the anatomical and functional distinction between memory processes as well as neural correlates and functioning of STM and LTM subsystems. In light of these findings, several memory models have been postulated.[4][5] 

While certain authors suggested the existence of a single memory system encompassing both short- and long-term storage,[6] after 50 years the Atkinson and Shiffrin model remains a valid approach for an explanation of the memory dynamics.

In light of more recent research, however, the model has several problems mostly concerning the characteristics of STM, the relationship between STM and WM as well as the transition from STM to LTM. 

Short-term memory: meaning and system(s)

It is a storage system that includes several subsystems with limited capacity. Rather than being a limitation, this restriction is an evolutionary survival advantage, since it allows paying attention to limited but essential information, excluding confounding factors. 

It is the classic example of the prey that must focus on the hostile environment to recognize a possible attack by the predator. Given the functional peculiarities of the STM (collection of sensorial information), the subsystems are closely related to the modalities of sensory memory.

As a consequence, there have been several sensorial-associated subsystems postulated, including the visuospatial, phonological (auditory-verbal), tactile, and olfactory domains. 

These subsystems involve different patterns and functional interconnections with the corresponding cortical and subcortical areas and centers. 

The concept of working memory

In 1974, Baddeley and Hitch developed an alternative model of STM which they termed as working memory.[7] Indeed, the WM model does not exclude the modal model but enriches its contents.

On the other side, the short-term store can be used to characterize the functioning of the WM. WM refers more to the entire theoretical framework of the structures and processes used for the storage and temporary manipulation of information, of which STM is only a component. In other words, STM is a functional storage element, while WM is a set of processes that also involve storage phases.

WM It is the memory that we constantly use, which is always “online” when we have to understand something or solve a problem or make an argument, the cognitive strategies for achieving short term goals.

The proof of the importance of this sort of ‘operating system’ of memory shows by the evidence that WM deficits are associated with several developmental disorders of learning, including attention-deficit hyperactivity disorder (ADHD), dyslexia, and specific language impairment (SLI).[8]

Short-term and Long-term memory

These types of memory can be classically distinguished based on storage capacity and duration. The capacity of the STM, indeed, has limitations in the amount and duration of information it can maintain. In contrast, LTM features a seemingly unlimited capacity that can last years. 

The functional distinctions between systems of memory storing and the exact mechanisms for how memories transfer from ST to LTM remain a controversial issue. Do STM and LTM represent one or more systems with specific subsystems?

Although the STM probably represents a sub-structure of the LTM, which is a sort of long-term activated storage, rather than looking for a ‘physical’ division, it seems appropriate to verify the mechanisms of transition from a memory that is only a passage to a lasting memory.

Although the classic multi-modal model proposed that storage of ST memories occurs automatically without manipulation, the matter seems to be more involved.

The phenomenon concerns quantitative (number of memories) and qualitative (quality of memory) features.

Regarding quantitative data, although the number of Miller of 7 +/- 2 items identifies the number of elements included among individual slots, the grouping of memory bits into larger chunks (chunking) could allow storing a lot more information of bigger size and continuing to keep the magic number. The qualitative issue, or memory modulation within processing, is a fascinating phenomenon.

It seems that the elements of STM undergo processing, which provides a sort of editing that involves the fragmentation of each element (chunking) and its re-elaboration and re-elaboration. This phase of memory processing is called encoding and can condition subsequent processing, including storage, and retrieval.

 The encoding process encompasses automatic (without conscious awareness) and effortful processing (through attention, practice, and thought) and allows us to retrieve information to be used to make decisions, answer questions, and so on.

There are three pathways followed during the encoding step: the visual (information represented as a picture), acoustic (information represented as a sound), and semantic encoding (the meaning of the information).

The processes interconnect with each other, so that information is broken down into different components. During recovery, the pathway that has produced the coding facilitates the recovery of the other components through a singular chain reaction. 

A particular perfume, for instance, makes us recall a specific episode or image. Of note, the encoding process affects the recovery, but the recovery itself undergoes a series of potential changes that can alter the initial content.

In neurofunctional terms, the difference between STM and LTM is the occurrence, in the LTM, of a series of events that must fix the engram(s) definitively. 

This effect occurs through the establishment of neural networks and expresses as neurofunctional phenomena including the long term potentiation (LTP) which is an increase in the strength of the neural transmission deriving from the strengthening of synaptic connections.

This process requires gene expression and the synthesis of new proteins and is related to long-lasting structural alterations in the synapses (synaptic consolidation) of the brain areas involved such as the hippocampus is the case of declarative memories.

The role of the hippocampal network

Of note, the hippocampal neurogenesis regulates the maintenance of LTP.[9] However, the hippocampal network, including the parahippocampal gyrus, hippocampus, and neocortical areas is not the place where memories are stored, but it has a crucial role in forming new memories and in their subsequent reactivation.

It seems that the hippocampus has a limited capacity and acquires information quickly and automatically without keeping it for long. Over time, the originally available information becomes permanent in other brain structures (in the cortex), independently from the activity of the hippocampus itself.

The crucial mechanism of this transfer is the reactivation (“replay”) of the configurations of neural activity. In other words, the hippocampus and the medial temporal structures connected to it are crucial for holding an event as a whole as it distributes in an organized way memory traces. It is an operating system that through different software can store, organize, process, and recover hardware files.

This hippocampal-guided reactivation (retrieval) leads to the creation of direct connections between the cortical traces and then to the formation of an integrated representation in the neocortex including the visual association cortex for visual memory, the temporal cortex for auditory memory, and the left lateral temporal cortex for knowledge of word meaning. Moreover, the hippocampus has other specific tasks, for example, in the spatial memory organization.

Other brain areas are involved in memory processes; for example, the learning of motor skills has links to the activation of the cerebellar regions and brainstem nuclei.

Furthermore, learning of perceptive activities (improvements in the processing of perceptive stimuli essential in everyday life activities such as understand spoken and written language)  involves, basal ganglia and sensory and associative cortices whereas learning cognitive skills (related to problem-solving) involve the medial temporal lobes initially.

More information: Memory retrieval modulates spatial tuning of single neurons in the human entorhinal cortex, Nature Neuroscience (2019). DOI: 10.1038/s41593-019-0523-z ,

Journal information: Nature Neuroscience
Provided by Columbia University School of Engineering and Applied Science


Please enter your comment!
Please enter your name here

Questo sito usa Akismet per ridurre lo spam. Scopri come i tuoi dati vengono elaborati.