A new study by neuroscientists at the Massachusetts Institute of Technology sheds light on how the brain creates internal maps of space. Their research in mice reveals that while some brain cells quickly encode specific locations, it takes a broader ensemble of neurons and repeated experiences—along with sleep—to form a coherent mental map of the environment.
The study, published in Cell Reports, supports the idea that cognitive maps are built through a gradual process involving not just specialized “place cells” in the hippocampus, but also a group of neurons that initially respond only weakly to specific locations. Over several days of exploration and sleep, these weakly tuned neurons begin to work together with place cells, forming coordinated patterns that reflect the layout of an environment.
The researchers were interested in a longstanding question in neuroscience: How does the brain go from recognizing individual places to constructing a complete internal map? Since the 1970s, scientists have known that certain hippocampal neurons fire when an animal is in a specific location. But a map requires more than isolated waypoints—it needs a network that connects them. Psychologist Edward Tolman first proposed the idea of cognitive maps in 1948, and while the discovery of place cells supported his theory, the exact process by which the brain links individual locations into a full map remained poorly understood.
“I’m interested in this project because how memory is formed in the brain is one of the most fundamental questions in neuroscience,” said study author Wei Guo, a research scientist at the Picower Institute of Learning and Memory, working under the supervision of Professor Matthew Wilson.
To investigate, the team used mice that freely explored unfamiliar mazes over several days. Importantly, the animals did not receive rewards or punishments in the mazes, allowing the researchers to study how the brain learns spatial layouts without reinforcement. The focus was on a form of passive learning known as latent learning, where knowledge is acquired without immediate behavioral changes.
To track brain activity, the researchers used advanced calcium imaging. They genetically modified hippocampal neurons to produce a fluorescent protein that signals activity, and implanted tiny lenses and microscopes into the mice’s brains. This setup allowed them to record activity from hundreds of neurons in the hippocampus while the mice explored the mazes or rested in their home cages.
Using a technique called manifold learning, the researchers created simplified visual representations of neural activity patterns over time. On the first day in a maze, each neuron had its own spatial firing pattern, but the ensemble of neurons did not yet form a recognizable map. By day five, however, the overall neural activity could be organized into a low-dimensional shape—called a “neural manifold”—that resembled the structure of the maze. This shift showed that over time, the brain began to represent the entire environment, not just individual spots.
The researchers found that this transformation was especially dependent on sleep. In one part of the study, mice explored the maze twice in one day, with a three-hour break in between. Some mice were allowed to sleep during the break, while others were gently kept awake. Only the sleeping mice showed improvements in how well their neural activity matched the maze layout. This indicated that sleep helped reorganize the hippocampal neural patterns into a more coherent map.
To understand what was changing in the brain, the researchers focused on two types of neurons. Some neurons, called strongly spatial cells, had clear place fields—they reliably fired when the mouse visited specific parts of the maze, even from the first session. These cells remained stable across days and did not significantly change their behavior during learning. In contrast, weakly spatial cells had less defined firing patterns early on but gradually increased their spatial tuning over time. More importantly, these weakly tuned cells became more coordinated with the rest of the neural network, particularly during sleep.
The researchers measured each neuron’s “mental field”—how its activity related not to physical space, but to the broader neural state. This helped them identify which neurons were becoming more integrated into the ensemble. They found that weakly spatial neurons, even if they never became strong place cells, played a key role in shaping the overall structure of the cognitive map. These cells developed correlations with other neurons, contributing to a network that could represent not just isolated locations but the relationships between them.
When the researchers tried to reconstruct the neural map using only the strong place cells, the resulting patterns showed little change over time. Only when they included the weakly spatial neurons did the full map-like structure emerge. This suggests that while place cells provide the building blocks, it’s the subtle and often-overlooked shifts in less specialized neurons that help assemble the full mental picture.
The study points to a broader role for weakly tuned neurons in learning. Rather than being noise or irrelevant, these cells help the brain build flexible, interconnected representations. They appear to respond not just to places, but to combinations of activity across the network. Over time, their activity becomes more synchronized with the rest of the ensemble, helping to stitch together a map of the environment.
“I was surprised that a subset of previously overlooked neurons that had weak activity turned out to be pivotal to memory formation,” Guo told PsyPost.
Sleep appeared to amplify this process. During rest after maze exploration, neural activity patterns resembled those seen during navigation, a phenomenon known as replay. This replay likely helps the brain reinforce connections between different places. The researchers found that after sleep, neural states during rest were more similar to those during maze runs, suggesting that the brain was using sleep to reinforce and refine the map.
This work supports the idea that memory formation is not limited to fast, discrete events. It often involves slower, distributed changes that rely on experience and sleep. The findings also highlight that “sleep is very important in transforming your experience into memory,” Guo said.
As with all research, the study has some limitations. The researchers relied on calcium imaging, which offers slower and less precise readings than direct electrical recordings. Their recordings were also limited to one part of the hippocampus and did not include other brain areas that may contribute to spatial memory.
Looking forward, Guo plans “to further investigate the local circuit in the hippocampus as well as their interactions with other brain regions during memory formation.”
The study, “Latent learning drives sleep-dependent plasticity in distinct CA1 subpopulations,” was authored by Wei Guo, Jie J. Zhang, Jonathan P. Newman, and Matthew A. Wilson.
