Neuroscientists at MIT's Picower Institute for Learning and Memory conducted a pioneering study that sheds new light on the brain's creation of unified cognitive maps of geography, highlighting the crucial role of sleep in this process. The findings, published in Cell Reports, provide a better understanding of the brain mechanisms that convert discrete location memories into a coherent spatial framework.

For decades, researchers have known that the hippocampus, a vital area of the brain, includes neurones known as "place cells" that activate when an individual is in a specific location. Although these neurones enable us to identify distinct locations, our understanding of the brain's capacity to form a continuous cognitive map is lacking. The notion of cognitive maps, developed in 1948, outlines how the brain conceptualizes landscapes as interconnected rather than separate points.

According to a new MIT study, cognitive maps evolve over several days through small alterations in the activity of "weakly spatial" cells rather than the robust activity of place cells. These cells, which were initially weakly associated with specific locations, serve an important role in connecting disparate regions into a unified brain map.

"On day one, the brain does not represent the space very well," said Wei Guo, the study's principal author. "But by day 5, the neurones work in a coordinated ensemble, forming a map of the space."

Latent learning is the form of learning that occurs without immediate reinforcement or explicit instructions. Unlike traditional types of learning, which rely on rewards or punishments, latent learning occurs naturally via observation and experimentation. Often, this process is subtle, as newly acquired knowledge or abilities remain dormant until a situation demands their application. The MIT study observed latent learning in mice as they freely explored mazes without explicit tasks or incentives, gradually internalizing the maze structure over time.

This phenomenon has far-reaching consequences for understanding how humans and animals navigate their environments. For example, a person wandering through a new neighborhood may instinctively notice landmarks, such as businesses, parks, and crossroads.  Even if there is no immediate necessity to recollect this information, it will become available later when planning a path or orienting themselves. The brain's ability to undertake latent learning demonstrates its extraordinary capacity to absorb and organize information passively, laying the groundwork for adaptive behavior.

Over several days, the researchers examined mice as they explored basic mazes of various shapes for 30 minutes each. Without any external incentives or rewards, the mice engaged in "latent learning," a process in which they gradually assimilated spatial layouts. We detected neural activity in the hippocampus's CA1 area during both exploration and subsequent sleep.

The study used a cutting-edge technique termed "manifold learning" to examine the activity of weakly spatial cells. Over time, these cells began to associate their activity patterns with overall brain network activity rather than specific regions. This interaction allowed the hippocampus network to encode a cognitive map that resembled the maze's architecture. "Weakly spatial cells" functioned as bridges, connecting discrete locations represented by place cells.

Sleep was crucial for this brain change. Allowing mice to sleep in between exploratory periods significantly refined their cognitive maps. Conversely, mice who did not receive enough sleep did not exhibit the same improvement.

"Sleep facilitates the consolidation and refinement of memories," stated Matthew Wilson, the study's principal author. During sleep, the neuronal "replay" of previous experiences allows the hippocampus to combine fragmented memories into cohesive maps. This process improves the cells' ability to encode bigger "mental fields" of network activity, as well as their spatial precision.

The maps created in the mice's brains were not accurate reproductions of the mazes but rather schematic approximations. These topological frameworks facilitate mental navigation without the necessity for physical presence. For example, once a person has created a cognitive map of a neighborhood, they can mentally plan routes or activities, such as having breakfast at a bakery followed by a stroll through a nearby park.

Wilson proposes that weakly spatial cells may also encode non-spatial information, such as emotional or contextual significance, enhancing cognitive maps with meaning. Despite the absence of landmarks and behavioural reactions in this study, it establishes the foundation for future investigations into the integration of such information into spatial representations.

The findings emphasize an important component of human learning: the brain's ability to build meaningful representations through spontaneous exploration and subsequent sleep. This type of unsupervised learning, which occurs without direct reward, is a key component of cognitive flexibility and intelligence.

"Substantial neuroplastic changes occur at the ensemble level during natural behavior and sleep," the researchers found. These findings may help shape techniques for improving memory and learning, especially in situations involving spatial navigation or abstract conceptualization.