What Is a Cognitive Map and How Does Your Brain Use It?

A cognitive map is your brain’s internal model of the space around you, built not from a single snapshot but from accumulated experience of landmarks, distances, and directions. The psychologist Edward Tolman coined the term in 1948 after watching rats navigate mazes in ways that suggested they had learned the layout of the space rather than just memorized a sequence of turns. Decades later, neuroscientists discovered the cellular machinery behind this idea: specialized neurons in the hippocampus and surrounding cortex that fire in patterns corresponding to where you are, which direction you face, and how far you’ve traveled. What makes the concept especially interesting is that this same neural architecture appears to organize non-spatial information too, from the passage of time to abstract relationships between ideas.

The Neurons That Draw Your Internal Map

The best-understood building block of the cognitive map is the place cell. These are hippocampal neurons that fire when an animal occupies a specific location in an environment and go quiet everywhere else.¹ Different place cells cover different spots, so collectively the population tiles the entire available space. When a rat enters a room it has explored before, the same constellation of place cells reactivates, essentially pulling up a stored map.

Place cells do not work alone. In the entorhinal cortex, a region that feeds information into the hippocampus, grid cells fire at regularly spaced locations that form repeating triangular patterns across an environment. Border cells, found in the same region, fire near the walls or edges of a space. Head direction cells, found in several brain areas, act like an internal compass: each one fires when the animal faces a particular direction, regardless of where it is.² Together, these four cell types give the brain a coordinate system: “where am I,” “which way am I facing,” “how far have I gone,” and “where are the boundaries.”

Grid cells deserve special attention because their hexagonal firing pattern functions as a kind of ruler for the brain. The spacing and orientation of the grid shifts in response to changes in the environment, including social cues. When a social landmark like another animal replaces a simple object, grid cells remap, and firing fields closer to the social landmark become more active.³ The map is not just geometric; it is flavored by what matters in the environment.

Building the Map From Sensory Experience

Your brain constructs a cognitive map by blending two streams of information. One is path integration, sometimes called dead reckoning: the brain tracks how far and in what direction you’ve moved using self-motion signals from muscles, the vestibular system, and optic flow. The other stream comes from external landmarks, the visual, auditory, and even olfactory features of the environment that anchor the map to the real world.

Smell, for instance, is a surprisingly potent anchor. Experiments have shown that odor cues enrich place cell representations and dramatically improve navigation. Presenting the same scent at different locations generates distinct place cell activity patterns, and a single odor landmark enhances local place cell density while also prompting the formation of new place cells beyond the cue itself.⁴ The finding underlines that cognitive maps are multimodal: they pull from every useful sensory channel, not just vision.

When path integration signals and visual landmarks conflict, the brain has to reconcile them. Research manipulating visual gain in virtual environments, so that the visual scene moves faster or slower than the animal’s actual movement, found that grid cells remapped significantly during both gain increases and decreases.⁵ The map destabilizes and then recalibrates. This is part of why virtual reality can sometimes feel disorienting: the normal agreement between what your body senses and what your eyes see is disrupted, and the neural map momentarily comes apart.

Switching Between Your Point of View and a Bird’s-Eye View

One of the subtler things the cognitive map system does is translate between two fundamentally different ways of representing space. Your immediate sensory experience is egocentric, everything is relative to your body, what is to your left, what is ahead. But the hippocampal map is largely allocentric, organized in world-centered coordinates independent of which way you happen to be facing.

The conversion between these frames happens along a gradient running from parietal cortex to retrosplenial cortex. Parietal regions handle more egocentric and local spatial information, while retrosplenial cortex specializes in global allocentric frameworks.⁶ It is not a clean handoff from one region to the other. Both areas contain a mix of egocentric and allocentric signals, and the transformation unfolds continuously as sensory input from primary cortical areas gets reshaped into the world-centered format the hippocampus uses.⁷

This reference-frame machinery is part of what allows you to give someone directions (“go two blocks north and turn left”) rather than just retrace your own exact steps. You can mentally rotate the map to take someone else’s perspective, or plan a novel route you have never walked, because the underlying representation is detached from your personal viewpoint.

Maps at Multiple Scales

Not all place cells paint with the same brush. Along the length of the hippocampus, place fields vary in size. Cells in the dorsal (roughly, the top) part of the hippocampus have small, precise fields suited for fine-grained local navigation. Toward the ventral (bottom) end, fields become larger and fuzzier, useful for representing broad spatial context. The simple story has been that this gradient maps neatly onto a zoom function: dorsal for close-up, ventral for wide-angle.

That clean story has become more complicated. Recordings in dorsal hippocampus have revealed individual cells with multiple place fields spanning a wide range of sizes, showing that multi-scale coding is not just a ventral specialty but exists throughout the hippocampal axis.⁸ The dense web of connections running along the hippocampus likely links place fields at all scales, allowing the brain to simultaneously track your precise location within a room and your broader position within a building or neighborhood.

A Map That Rewrites Itself

Cognitive maps are not fixed diagrams etched into your neurons. They are flexible and get updated as the environment changes. When the shape of an enclosure is gradually morphed from square to round, individual place cells adjust their firing rates, sometimes up, sometimes down, and different fields of the same cell can shift independently.⁹ This rate remapping lets the brain register that the geometry of the space has changed while still recognizing it as the same basic environment.

In more dramatic situations, the map undergoes global remapping: the entire pattern of place cell firing rearranges as if the brain has loaded a completely different map. Walking into a new room triggers this. So does a sudden change in context, like the lights going off or a strong new odor flooding the space. The brain maintains a library of maps and switches between them depending on the situation. This is part of why you can navigate your home in the dark and your office under fluorescent lights without confusing the two: different contexts call up different stored maps.

Memory consolidation plays a role in stabilizing these maps over time. During sleep and quiet rest, the hippocampus produces sharp wave-ripple events, bursts of activity in which place cells replay sequences representing past or potential future experience. These replays support both the consolidation of spatial memories and the retrieval of stored maps for decision-making.¹⁰ In a real sense, your brain rehearses its maps while you sleep.

Mapping Things That Are Not Places

One of the most striking findings of the last decade is that the cognitive map system does not only handle physical space. The hippocampus and entorhinal cortex appear to use the same architecture to organize abstract relationships. Research on navigation tasks has shown that these regions encode not just the spatial distance to a goal but also a more direct, as-the-crow-flies distance that reflects an abstract relationship between the navigator and the target. The authors of one study noted that the brain may use the same system it evolved for navigating continuous space to represent the relationships among discrete entities in an abstract space.¹¹

Time is another dimension the hippocampus maps. Neurons called time cells fire at specific moments during a temporal gap, tiling the delay period the way place cells tile space. Even after accounting for the animal’s location, speed, and direction, these cells show robust temporal coding that peaks at specific moments and is organized sequentially across the population.¹² The implication is profound: the hippocampus does not just know where you are, it knows when you are in a sequence of events. This likely contributes to episodic memory, your ability to remember not just what happened but the order in which events unfolded.

Recent work has also found a mathematical connection between spatial navigation and language processing, suggesting that the computational principles the brain uses to move through physical space may also support how you move through a sentence or a narrative.¹³ The idea that spatial cognition scaffolds more abstract forms of thought is still being explored, but it fits neatly with the observation that we talk about ideas in spatial metaphors (“close to the truth,” “a distant memory”) without thinking twice.

Evidence From the Human Brain

Most of what we know about place cells and grid cells comes from electrodes implanted in rodent brains. A reasonable question is whether humans have the same system. The evidence, gathered through functional MRI rather than single-cell recordings, says yes.

In humans, the hallmark of grid cell activity is a signal in entorhinal cortex that shows six-fold rotational symmetry: brain activity fluctuates with a 60-degree periodicity as a function of movement direction, exactly what you would expect from a population of grid cells firing in a hexagonal pattern. This signal has been detected during virtual navigation in an open field, appearing specifically during movement and not during stationary periods.¹⁴ No other rotational symmetries, like four-fold or eight-fold, showed up, ruling out a generic directional signal and pointing specifically to grid cells.

Even more surprisingly, this six-fold signal appears during imagined navigation. When people close their eyes and mentally move through a familiar virtual environment, entorhinal cortex produces the same grid-like pattern seen during actual movement.¹⁵ The cognitive map does not require sensory input to activate; it can run as a simulation. This probably explains how you can mentally walk through your apartment to count the windows or plan a route to a new restaurant by picturing the streets.

Researchers have also looked for three-dimensional grid signals in humans navigating vertical space. An fMRI study found evidence consistent with a face-centered cubic grid model in left entorhinal cortex, with voxel orientations clustered around a coherent mean direction.¹⁶ While the evidence for three-dimensional mapping in humans is still emerging, single-cell recordings in bats, animals that navigate in true three dimensions, have found place cells encoding three-dimensional volumes with similar resolution along all three axes.¹⁷ Most studies of spatial neurons have used flat environments, but the natural world is hilly and irregular, and the mapping system appears equipped to handle that.¹⁸

When the Map Starts to Crumble

Getting lost is one of the earliest symptoms of Alzheimer’s disease, and the cognitive map system is one of the first casualties. The entorhinal cortex, where grid cells reside, is among the first brain regions to develop Alzheimer’s pathology. In mouse models of the disease, grid cells become spatially unstable, particularly toward the center of an environment away from boundary cues. These unstable grid cells also show impaired integration of self-motion information, measured by reduced theta phase precession, a rhythmic firing pattern tied to distance tracking.¹⁹ The grid cells can still register the walls and borders of the space, but they lose their ability to track how far the animal has traveled, which explains why people with early Alzheimer’s can recognize familiar landmarks but still get lost between them.

The timeline of grid cell decline tracks closely with the timeline of behavioral deficits. In transgenic mice carrying Alzheimer’s-related mutations, the severity of grid cell disruption parallels worsening performance on path integration tasks as the animals age.²⁰ When the protein tau accumulates in the entorhinal cortex, the resulting excitatory cell loss leads to grid cell dysfunction and spatial memory deficits that mirror the disorientation seen in early human Alzheimer’s.²¹ Navigation tests are now being explored as potential early screening tools for the disease, since spatial difficulties can precede memory complaints by years.

Alzheimer’s is not the only condition that affects cognitive maps. People with schizophrenia appear to form what researchers describe as “shallow” cognitive maps. In a structured task, individuals with schizophrenia relied more on simpler visual features and less on abstract, higher-order features relevant to planning. Interestingly, they could still perform the task at a level comparable to controls, suggesting that the map was degraded but compensated for through other cognitive strategies.²² The distinction between forming a rich map and using it effectively for goal-directed behavior turns out to be a meaningful one for understanding cognitive deficits across psychiatric conditions.

Machines That Navigate Like Mammals

The discovery of grid cells did not just change neuroscience; it influenced artificial intelligence. A team at DeepMind trained a recurrent neural network to perform path integration, and the network spontaneously developed internal representations resembling grid cells, along with patterns similar to other entorhinal cell types. When this grid-like representation was used as the spatial backbone for a deep reinforcement learning agent, the agent could locate goals in unfamiliar and changing environments, outperforming both a human expert and comparison agents that lacked the grid structure. The grid-equipped agents could even take novel shortcuts, a behavior characteristic of animals with intact cognitive maps.²³

The emergence of grid-like patterns in an artificial system trained purely on a navigational objective, with no instruction to mimic biology, suggests that the hexagonal grid is not an evolutionary accident but something close to an optimal solution for encoding metric space. This has practical implications for robotics: autonomous vehicles and drones navigating GPS-denied environments could benefit from algorithms modeled on entorhinal grid coding, and several research groups are now pursuing that line of work. It is one of those cases where understanding the biology feeds directly back into engineering, and where the engineering success lends confidence that the biological theory is on the right track.