You see a woman on the street who looks familiar—but you can’t remember how you know her. Your brain cannot attach any previous experiences to this person. Hours later, you suddenly recall the party at a friend’s house where you met her, and you realize who she is.
In a new study in mice, researchers have discovered the place in the brain that is responsible for both types of familiarity—vague recognition and complete recollection. Both, moreover, are represented by two distinct neural codes. The findings, which appeared on February 20 in Neuron, showcase the use of advanced computer algorithms to understand how the brain encodes concepts such as social novelty and individual identity, says study co-author Steven Siegelbaum, a neuroscientist at the Mortimer B. Zuckerman Mind Brain Behavior Institute at Columbia University.
The brain’s signature for strangers turns out to be simpler than the one used for old friends—which makes sense, Siegelbaum says, given the vastly different memory requirements for the two relationships. “Where you were, what you were doing, when you were doing it, who else [was there]—the memory of a familiar individual is a much richer memory,” Siegelbaum says. “If you’re meeting a stranger, there’s nothing to recollect.”
On supporting science journalism
If you're enjoying this article, consider supporting our award-winning journalism by subscribing. By purchasing a subscription you are helping to ensure the future of impactful stories about the discoveries and ideas shaping our world today.
The action occurs in a small sliver of a brain region called the hippocampus, known for its importance in forming memories. The sliver in question, known as CA2, seems to specialize in a certain kind of memory used to recall relationships. “[The new work] really emphasizes the importance of this brain area to social processing,” at least in mice, says Serena Dudek, a neuroscientist at the National Institute of Environmental Health Sciences, who was not involved in the study.
A student in Siegelbaum’s lab helped establish this role about a decade ago, when he developed a genetic method for silencing CA2 in mice. With CA2 out of commission, mice could no longer tell an unfamiliar mouse from a littermate. The memory deficit was limited to social contexts as opposed to, say, memory for objects or locations that had no social meaning: the mice could still recognize familiar objects, for instance, or navigate a maze. “The surprise was that this one particular subregion was so critical in the mice for social memory,” Siegelbaum says.
But it was still not clear how the cells in the region were performing this function. “The question was: What’s going on in the brain of these mice in the CA2 region?” Siegelbaum says. To answer this question, Siegelbaum’s team had to devise a way to record neural activity in CA2 during social interactions—and to analyze that activity.
In 2018 then graduate student Lara Boyle began tackling the problem of recording the goings-on in CA2 using a miniscope, a type of microscope small enough to be worn like a hat on a mouse’s head. Boyle, a co-author of the new paper, endowed CA2 neurons with a protein that glows in the presence of calcium, the mineral that rushes into neurons when the cells are active. She positioned the microscope to detect the glow, measure its intensity and convert the measurements to electrical signals to be processed by a computer.
The apparatus recorded the activity of 50 to 60 neurons at a time as a mouse interacted with either two unfamiliar mice, two littermates or a littermate and a stranger. The mice in each of these pairs were contained separately in small wire “cups” on the left and right sides of the cage. Comparing the brain activity during the interactions, the researchers hoped, would reveal how the mouse recognized other mice as strangers or littermates or distinguished them as individuals. As it happened, however, the scientists could not make enough sense of the signals at first to determine how the brain was making these calls.
So in 2020 Boyle and Siegelbaum teamed up with Zuckerman neuroscientist Stefano Fusi and postdoctoral fellow Lorenzo Posani, who built a “linear decoder,” software that could decipher the glut of neural patterns. As reported in the new paper, the decoder processed the responses of individual mice to ferret out how the brain encrypts social familiarity and identity. The mouse brain uses a “very special code” for representing other mice, Fusi says.
The decoder uncovered a neural signature for the concepts of novelty and familiarity that applied across various pairs of both newly encountered and familiar mice. “A [proverbial] light would shine somewhere in the brain that says ’novelty’ or ‘familiarity,’ independent of the identities of these mice,” Siegelbaum says. “That was probably one of the biggest aha! moments of the study.”
What’s more, the better a mouse’s CA2 neurons distinguished between a novel and a familiar animal, the more adept the animal seemed to be at telling the mice in the cups apart. Because mice like novelty, they prefer to spend time with a stranger rather than a littermate. So the animals whose brains had the sharpest novelty detectors, as identified by the linear decoder, spent the most time sniffing the novel mouse relative to the familiar one.
The researchers also found neural patterns that enabled the mice to distinguish two familiar animals or two strangers. “You can decode with some level above chance not just whether [a mouse is perceiving] a novel animal or a familiar animal but [also] the identity of that animal,” Dudek says.
And the way the brain captures that identity is different for strangers than it is for familiar animals, Siegelbaum says. The researchers determined, for example, that the neural code for social identity depended on the location of the mice in the cups to a greater degree when those mice were familiar, consistent with the idea that memories of known individuals are rooted in a place, among other details. By contrast, no such details are attached to strangers, so the code is simpler, he says.
The principles behind these newly discovered brain signatures could inform better machine-learning systems, Fusi says. At present, computers must be trained on new information in very specific ways, or they suffer “catastrophic forgetting” of previous knowledge. “If you want [a machine-learning system] to continually learn for an entire life, we don’t have a way of doing that,” Fusi says. “Machines don’t learn in a natural environment like we do.” The new understanding of how the animal brain encodes social information, he says, could spawn ideas for solving the problem of catastrophic forgetting.
The study’s results also represent a small step toward a complete understanding of a social memory, says Thomas McHugh, a neuroscientist at the RIKEN Center for Brain Science in Japan, who was not involved in the research. “What we really want to understand as a field is how a memory’s formed that integrates all these components—who we are interacting with, where we are, what we’re doing—the contents of an event or episode,” McHugh says. “This [study] gives us some thoughts on how we might do that.”
CA2 is unlikely to act alone in this process because it connects to other brain regions that also play roles in social learning, memory and behavior, Dudek says. But if the findings in mice apply to humans, the work could help researchers uncover the roots of social difficulties in people, such as those that occur in schizophrenia and autism. Scientists could, for example, look for changes in the neural code of CA2 neurons in genetic mouse models of these conditions. “It’s important to have this baseline data to understand what changes in these models” as a step to understanding what changes in humans, McHugh says.
From there, researchers might find ways to normalize that circuitry to improve social memory, Siegelbaum says. “How many different disorders of memory are there?” he adds. “Do they all involve the same type of changes in neural processing, or are there more specific changes that are associated with different forms of disease?” Having a better way to classify these changes may lead to more targeted treatments, Siegelbaum says.