No matter how he looked at the data, Albert Tsao couldn’t see a pattern. Over several weeks in 2007 and again in 2008, the 19-year-old undergrad trained rats to explore a small trial arena, chucking them pieces of tasty chocolate cereal by way of encouragement. He then recorded the activity of individual neurons in the animals’ brains as they scampered, one at a time, about that same arena. He hoped that the experiment would offer clues as to how the rats’ brains were forming memories, but “the data that it gave us was confusing,” he says. There wasn’t any obvious pattern to the animals’ neural output at all.
Then enrolled at Harvey Mudd College in California, Tsao was doing the project as part of a summer internship at the Kavli Institute for Systems Neuroscience in Norway, in a lab that focused on episodic memory—the type of long-term memory that allows humans and other mammals to recall personal experiences (or episodes), such as going on a first date or spending several minutes searching for chocolate. Neuroscientists suspected that the brain organizes these millions of episodes partly according to where they took place. The Kavli Institute’s Edvard Moser and May-Britt Moser had recently made a breakthrough with the discovery of “grid cells,” neurons that generate a virtual spatial map of an area, firing whenever the animal crosses the part of the map that that cell represents. These cells, the Mosers reported, were situated in a region of rats’ brains called the medial entorhinal cortex (MEC) that projects many of its neurons into the hippocampus, the center of episodic memory formation.
Inspired by the findings, Tsao had opted to study a region right next to the MEC called the lateral entorhinal cortex (LEC), which also feeds into the hippocampus. If the MEC provided spatial information during memory formation, he and others had reasoned, maybe the LEC provided something else, such as information about the content of the experience itself. Tsao had been alternating the color of the arena’s walls between trials, from black to white and back again, to see if LEC neurons showed consistently different firing patterns in each case. But he was coming up empty-handed.
While Tsao struggled to make sense of his data, a researcher on the other side of the Atlantic Ocean was tackling a seemingly un-related problem. Marc Howard, a theoretical and computational neuroscientist then at Syracuse University, had filled a chalkboard with equations describing how the brain might achieve the complex task of organizing memories, according not to where they were formed, but to when. His mathematical model showed that if the passing of time was represented in a certain way in neural circuits, then that time signal could be converted into a series of mental “time stamps” during memory formation to help the brain organize past experiences in chronological order. Without data to confirm his model, however, the idea remained just that: an idea.
It would be several years before the two researchers became aware of each other’s work. By the time they did, neuroscientists had started thinking in new ways about how the brain keeps track of when experiences occurred. Today, the theoretical and experimental advances made by Howard, Tsao, and others in this field are helping to reshape researchers’ understanding of how episodic memories are formed, and how they might influence our perception of the past and future.
Back in 2008, however, Tsao was focused on finishing college. When his second summer in Norway came to an end, he left the Kavli Institute and his confusing dataset behind, and returned to California.
Another dimension
When the cognitive neuroscientist Endel Tulving coined the term “episodic memory” in a book chapter in 1972, he observed that recalling the content of memories was linked to a strong subjective sense of where and when an episode took place. The where component has been a focus of neuroscientific research for decades. In 1971, University of College London neuroscientist John O’Keefe discovered place cells, neurons in the hippocampus that fire in response to an animal being in specific locations. He shared the Nobel Prize in Physiology or Medicine with the Mosers in 2014 for their discovery of grid cells in the MEC, and several studies published since suggest that grid cells help the hippocampus generate place cells during memory formation.
How the brain encodes the when of memories has received far less attention, notes Andy Lee, a cognitive neuroscientist at the University of Toronto. “Space is something we see, it’s easy to manipulate… . It’s somewhat easier for us to grasp intuitively,” he says. “Time is much harder to study.”
Despite the thorniness of the subject, researchers have established in the last decade or so “that the brain has multiple ways to tell time,” says Dean Buonomano, a behavioral neuroscientist at the University of California, Los Angeles, and author of the 2017 book Your Brain is a Time Machine. Time is integral to many biological phenomena, from circadian rhythms to speech perception to motor control or any other task involving prediction, Buonomano adds.
One of the biggest breakthroughs in understanding time as it relates to episodic memory came a few years after Tsao completed his internship, when the late Boston University neuroscientist Howard Eichenbaum and colleagues published evidence of “time cells” in the hippocampus of rats. Hints of time-sensitive cells in the hippocampus had been trickling out of labs for a couple of years, but Eichenbaum’s study showed definitively that certain cells fire in sequence at specific timepoints during behavioral tasks: a rat trained to associate a stimulus with a subsequent reward would have one hippocampal neuron that peaked in activity a couple hundred milliseconds after the stimulus was presented, another that peaked in activity a few hundred milliseconds after that, and so on—as if the hippocampus were somehow marking the passage of time.
The findings, which are beginning to be extended to humans thanks to work by Lee’s group and a separate team at the University of Texas Southwestern, among others, generated interest in the representation of time alongside space in episodic memories. Yet it was unclear what was telling these cells when to fire, or what role, if any, they played in the representation of time passing within and between individual episodic memories. For Marc Howard, long fascinated by questions about the physical nature of time and the brain’s perception of it, the puzzle was a captivating one.
In the years leading up to Eichenbaum’s paper, Howard and his postdoc Karthik Shankar had been developing a mathematical model based on the idea that the brain could create a proxy for the passage of time using a population of “temporal context cells” that gradually changes its activity. According to this model, all neurons in this population become active following some input (a sensory stimulus, for example), and then relax, one by one, creating a gradually decaying signal that is unique from moment to moment. Then, during memory formation, the brain converts this signal into a series of sequentially firing “timing cells,” which log moments within a memory. The same framework could also work to tag entire episodes according to the order in which they took place.
The specific mathematical details of the model—in particular, the use of an operation called a Laplace transform to describe how temporal context cells compute time, and the inversion of that transform to describe the behavior of the hypothesized timing cells—nicely recapitulated several known features of episodic memory, such as the fact that it’s easier to remember things that happened more recently than things that happened a long time ago. And after hippocampal time cells, with their sequential firing patterns, were described in 2011, Howard, by then at Boston University, was gratified to see that they seemed to possess many of the properties he and Shankar had predicted for their so-called timing cells.
But the first piece of the puzzle was still missing. No one had identified the gradually evolving set of temporal context neurons needed to produce the time signal in the first place, Howard says. “We waited a long time for somebody to do the experiment—really just moving the electrodes over to the LEC and looking for it.”
Finding a signal
After graduating from Harvey Mudd in 2009, Tsao returned to the Kavli Institute for a PhD. Although he mostly worked on other projects, by the end of his program he’d convinced himself, and the Mosers, that the rat experiments from his summer internship were worth another look. Tsao was “an exceptional student,” May-Britt Moser says, and the Kavli team trusted that his data were correct, but “we didn’t know what we were seeing.” The neurons in the LEC seemed to be behaving so unpredictably.
Digging back into his old work after he graduated from his PhD program, Tsao began thinking about better ways to analyze the dataset. “We had always looked at activity at the level of individual neurons,” he says. “At some point, we decided to look at it at the entire population level.” In doing so, Tsao revealed that LEC activity was, in fact, changing—gradually, within and between trials.
Data from further experiments, carried out by Kavli researchers after Tsao moved to Stanford University for a postdoc in 2015, showed that a whole cluster of cells within the LEC became active at the beginning of trials, and then that activity decayed as individual neurons relaxed at various rates. Other cells in the LEC, meanwhile, seemed to become gradually less (or sometimes more) active over the course of the entire experiment. Looking at the data this way, the team was able to distinguish individual trials not just according to wall color but, far more intriguingly, by the order in which the rat had done them, explains May-Britt Moser. “Together, [these cells] coded for time.”
Publishing the findings in late 2018, the team cited Howard’s and Shankar’s work, highlighting how the sort of activity patterns Tsao had seen in the LEC neuronal population matched up with the pair’s theoretical predictions. The Norwegian group also noted that this evolving signal seemed able to track passing time over multiple timescales—changing fast enough to distinguish between individual moments on the scale of seconds within a single episode, as well as to distinguish whole episodes from one another over the scale of minutes or hours. On reading the team’s findings, “I was ecstatic,” Howard says. “It was really a big deal for me.”
The paper was exciting for many in the neuroscience community, and its publication was followed by a burst of theoretical work from several groups, not just Howard’s. Edmund Rolls, a computational neuroscientist at the University of Warwick, incorporated the findings from the Kavli group’s 2018 paper into a model that explored how interacting networks in the brain might convert gradually changing LEC activity into a sequence of hippocampal time cells, based on a framework he’d developed more than a decade earlier to explain how grid cells might lead to the generation of place cells.
Additional experimental data started flowing in, too. Howard and colleagues, for example, analyzed recordings from monkeys’ entorhinal cortex—an area containing the MEC and LEC—and found activity similar to that observed in Tsao’s rats, according to a preprint published last summer on bioRxiv. Specifically, a cluster of neurons in the entorhinal cortex spiked after a monkey was presented with an image, and then returned to baseline, with different neurons relaxing at different rates. Just a couple of months later, researchers in Germany reported that activity recorded from the human LEC could be used to reconstruct the timeline of events people experienced during a learning task.
The gradual change in LEC activity wasn’t the only novel result from Tsao’s paper. Several groups picked up on a related finding that the rate of change in the LEC—and indeed in many areas of the brain—may depend on the sort of experience an animal is having. That phenomenon might help explain why the passage of time within episodic memories seems so subjective.
Personal time
As a follow-up to his original experiments with the rat arena, Tsao had done a couple of additional trials during his Kavli internship with a figure-eight maze. In each of those trials, instead of freely exploring an arena, the rat would run around the maze, following the track left, then right, then left, and so on. After discovering patterns in rats’ LEC neuronal firing during arena trials, Tsao hoped to see something similar in data from the figure-eight mazes—something that would distinguish trials from one another according to when they took place. “But … it turned out we couldn’t tell them apart very well,” says Tsao. “For a while this was very disappointing—this was basically the opposite conclusion that we had reached from the [arena] experiment.”
It wasn’t until Tsao dug into the literature on episodic memory that he came to realize what might be going on. “Maybe it’s not so much about physical time, as you measure in clocks, but more about subjective time, as you perceive it,” he says. Running in a twisted loop was a repetitive, boring task compared to exploring an arena, and the rat’s LEC seemed to reflect that by changing its activity less substantially during the figure-eight experiment than it had during the arena experiment. It seemed as though the rat’s brain wasn’t really experiencing individual figure-eight trials as distinct events, at least not to the extent it had for arena trials, Tsao says.
This link between the type of experience and the way time is represented in neurons touches on a well-known quirk of episodic memory. It’s easier to pick out memories from a week of exciting and varied activities than from a week filled with normal, uninteresting tasks, and the former feels much longer than the latter when it’s recalled. (This is different from the sensation of time dragging when doing something boring—an effect of consciously counting time as it passes rather than representing it in a memory of the event, notes Buonomano.) Tsao’s study hinted that part of this subjective effect might arise because the LEC, which receives neural input from areas involved in processing sensory information, changes its activity to a greater degree during more complex experiences than during ones that require little processing. It implies, Tsao speculates, that time in memory might be entirely “drawn from the content of your experiences, as opposed to being coded as an explicit thing.”
Although neural recordings are challenging to carry out in humans, functional MRI (fMRI) data from other research groups has helped flesh out the link between the rate of activity changes in the cortex and the representation of time in memory. Kareem Zaghloul, a neurosurgeon and neuroscientist at the National Institute of Neurological Disorders and Stroke and a “big fan of Marc Howard’s work and his model,” had been running an experiment on the effects of brain stimulation on human memory around the time Tsao’s paper came out. As part of their project, Zaghloul and his colleagues decided to use their dataset to look at how temporal context might influence memory formation. “We hypothesized that perhaps the extent to which these signals of time change, maybe that affects your ability to distinguish memories from one another,” Zaghloul says.
Participants in his group’s study had been asked to learn pairs of words, such as “pencil” and “barn,” and then remember these pairs later while avoiding confusing them with other pairs they’d learned, such as “orange” and “horse.” Measuring activity using electro-encephalography across broad regions of participants’ brains while they learned the word pairs, the researchers found that the faster a person’s neural activity changed during the learning task, the better they performed on memory recall later on. Electrical stimulation of participants’ brains during the learning task didn’t have a consistent effect on the rate of activity change, Zaghloul adds, “but when it made it faster, people tended to do better at remembering the word pairs, and when it made it slower … people tended to do worse,” he says. The findings, published last year, suggest “that this representation of time does play a role in your ability to lump or distinguish memories,” he says.
That the sense of time in episodic memory might be dependent on neural activity rather than on a traditional clock reinforces some researchers’ belief that the brain perceives time rather differently from how people imagine it to. Buonomano and New York University neuroscientist György Buzsáki have independently argued before and since Tsao’s work that neuroscientists should rely less on preconceived notions of time and instead think more about how time-related information might be used by the brain. “The sole function of memory is to allow animals to better prepare for the future,” says Buonomano. “Sometimes the field forgets that detail.”
Thinking ahead
Tsao is still studying the brains of rats as a postdoc at Stanford University, although his focus has shifted to other topics in neuroscience. But for other researchers studying the representation of time in episodic memory, the work has only just begun.
May-Britt Moser says her group is continuing the line of research Tsao started, exploring how the hippocampus in rats integrates temporal and spatial information from the LEC and MEC during memory formation. The idea’s been around for a while. Several years ago, Eichenbaum and colleagues reported that rats’ time cells seem sensitive to spatial as well as temporal information. More-recent research has complicated the story further, identifying time cells outside the hippocampus, and finding that some place cells seem to respond to time-related signals from the LEC, leading some neuroscientists to propose that the hippocampus possesses different time-tracking systems for different timescales.
To Howard, one of several theoreticians who has modeled how the brain might combine signals encoding the when and where of episodic memories, the blurred boundary between space and time is intuitive. Having originally trained in physics, he says, “I was pretty sure that the brain’s representation of space and the brain’s representation of time ought to obey the same equations,” whatever the scale. He and many other neuroscientists are now working under the assumption that the brain uses a unified representation of space and time in remembered experiences. And at least for some aspects of memory, Howard says, “I think that’s the story that’s starting to unfold now.”
While Tsao’s work focused on how time is encoded during memory formation, some groups are working on the other side of the coin: what happens during the process of memory retrieval. Researchers at the University of California, Irvine, recently reported that people who showed higher LEC activity during a memory retrieval task were better at recalling when specific events in a sequence happened, supporting a role for the LEC in a sense of time during memory retrieval as well as formation. Zaghloul, Howard, and others, meanwhile, have independently published work showing that when people successfully recall memories, they seem to reinstate the activity patterns in the medial temporal lobe—a region that includes the hippocampus and the entorhinal cortex—that were present when that memory was formed. It’s an effect, notes Zaghloul, that’s thought to allow a sort of “jump back in time” on recalling a memory.
Such an ability to reinstate past activity patterns could have applications to the brain’s representation of events that haven’t yet happened, too, Howard says. “It occurred to us quite a while ago that if the brain has equations of the past, you could construct an estimate of the future with the same types of properties.” Empirical data to test the idea are lacking for now. One of the first things to do will be to figure out how the brain could skip back or forward to different activity states, because “we don’t currently have algorithms that can do that,” Howard notes. “We’re actively working on … figuring out a set of equations to describe the how of jumping back in time. Actually, I’m looking at my chalkboard right now, and I’m pretty optimistic.”
To be successful as a social animal, you need to know where you stand relative to others. Brain cells that perform precisely this function–locating the ‘self’ and others in space–have now been identified. In rats, the same brain area that stores the animal’s own location also maps the movements of other rats. Sometimes these representations are processed jointly by the same cells, depending on a rat’s goals and actions. This discovery, from Japan’s RIKEN Brain Science Institute, deepens our understanding of the hippocampus and its role as the brain’s positioning system.
It has been known for some time that the hippocampus maintains a mental map of space–in fact, the 2014 Nobel Prize in Physiology or Medicine was awarded precisely for this research. ‘Place cells’ and ‘grid cells’ in the hippocampus register the location of the brain’s owner in its environment, but until now, little was known about how the movements of others are tracked in the brain. Researchers put this to the test by observing the activity of hippocampal neurons in one rat (the ‘self’) watching another rat (the ‘other’) go through a simple T-maze. The self’s neurons registered what the other was doing and changed their responses based on the self’s location and subsequent actions. This study was published on January 11 in Science, which also contains a report of similar location awareness in the brains of bats.
Hippocampal place cells light up only for certain locations in the environment, and some of these cells clearly preferred the location of the other rat. Besides activation to places, the timing of neural activity in the hippocampus is also important. A ‘refresh rate’ of around 8 Hz dictates how often neurons update their activity, a phenomenon called theta-cycle phase precession. Three-quarters of cells updated based on the other’s location, not only the self’s. “It is very interesting that the trajectories of the other, that is, the past, current, and future positions of the observed rat, are compressively represented in 100-millisecond cycles in the hippocampus,” says Shigeyoshi Fujisawa, research group leader at the RIKEN Brain Science Institute.
“Spatial representations of self and other in the hippocampus” by Teruko Danjo, Taro Toyoizumi, and Shigeyoshi Fujisawa in Science. Published online January 12 2018 doi:10.1126/science.aao3898
These are the four types of spatial models for the hippocampus that are proposed in this paper.NeuroscienceNews.com image is credited to RIKEN.
Researchers at the German Center for Neurodegenerative Disease (DZNE) have found a possible explanation for the difficulty in spatial orientation experienced sometimes by elderly people. In the brains of older adults, they detected an unstable activity in an area that is central for spatial navigation.
The research is in Current Biology. (full open access)
Two free examinations in rodents discover grid cells, which structure the mind’s map, are more plastic than thought and are liable to the impact of remuneration.
During the 1970s, researchers found that specific neurons in the hippocampus—a region of the cerebrum associated with learned and memory—would fire in light of specific areas. They were classified “place cells,” clarifies Charlotte…
Perhaps most tantalizing of all, an experiment conducted in 2016 introduced a far more abstract context for grid cell behavior. Researchers led by Timothy Behrens, a computational neuroscientist at the University of Oxford, had people watch the silhouette of a bird on a screen as the length of its neck, the length of its legs or both were stretched and compressed. The hexadirectional signal arose in their fMRI data, in several areas of the brain; it varied just as if the test subjects were navigating a two-dimensional “bird space,” where one axis denoted neck length, the other leg length.
The finding suggested that the brain processes trajectories through physical spaces and conceptual spaces in much the same way. Now, researchers including Behrens, Bellmund and neuroscientist Christian Doeller propose that all knowledge can be plotted this way, in terms of features of interest — that different objects, different experiences and different memories can be organized and traversed with the grid code.
“The Brain Maps Out Ideas and Memories Like Spaces“ from Quanta Magazine
The authors propose that we navigate our abstract thoughts using the same neurons as for navigating physical space. Therefore, perhaps one of the basic principles of abstract thought is that we develop “cognitive spaces” in which we arrange information, and then can mentally navigate around that information. This hypothesis has the strength of fitting with much of our subjective experience. For example, there is a phenomenon known as embodied cognition – abstract ideas are anchored to physical metaphors. We talk about a “big” idea, or a “deep” thought, someone can be hierarchically “above” someone else, and argument can be “strong”, “weak”, or even “beautiful”. We use physical metaphors about size, physical characteristics, and relationships to represent abstract ideas.
This may also explain the appeal of infographics, where information is represented graphically using physical and spacial relationships.
What the authors are suggesting is that perhaps our brains are wired specifically to do this – that we are literally navigating our abstract thoughts. Perhaps we don’t do this all the time, but this may be the way, or one way, in which we organize information in our brains. I even learned in a formal course to deliberately and consciously do this – to enhance memory by arranging information in a familiar physical space (a memory house or memory palace). This strategy works really well.
“Navigation May Be Fundamental to Thinking“ from NeuroLogica
