Abstract

Spatial orientation is the capacity to know one's position and heading in the environment and to keep track of them while moving. Cognitive psychology studies it through two intertwined problems: the frames of reference in which spatial information is coded — egocentric, centred on the body, and allocentric, centred on the world — and the neural machinery that builds and maintains an internal cognitive map. This article traces the field from Tolman's cognitive map through the discovery of place, head-direction, and grid cells to modern accounts of human wayfinding, path integration, and the wide individual differences in navigational skill. Three interactive demonstrations let a reader switch between reference frames, judge the mental rotation of a figure, and watch a homing vector accumulate along a path.

Keywords: spatial orientation, cognitive map, reference frames, path integration, grid cells

Spatial orientation is the ability to represent where one is, which way one is facing, and how the surrounding layout is arranged, and to update that representation continuously as the body moves. It is a precondition for almost every spatial act: reaching for a cup, crossing a room, or finding the way home across a city all depend on maintaining a stable sense of self-location against a shifting flow of sensory input. The problem is harder than it appears, because the raw signals — a retinal image, the pull of gravity, the sense of a turn — are all tied to the momentary pose of the body, yet the goal is often to act with respect to the world, which does not move when the observer does. Cognitive psychology has organised the study of spatial orientation around two questions that run through the rest of this article: in what frame of reference is spatial information held, and by what neural code is a map of space constructed and kept current. The first question distinguishes representations centred on the observer from those centred on the environment; the second traces a chain of specialised neurons — place cells, head-direction cells, grid cells, and their relatives — that together furnish the brain with something close to an internal positioning system.

Key Takeaways
  • Spatial orientation is the ongoing computation of self-location and heading, not a static snapshot of where things are.
  • Spatial information is coded in two broad frames of reference — egocentric, centred on the body, and allocentric, centred on the world — and flexible behaviour requires translating between them.
  • Tolman's cognitive map was given a physical basis by the discovery of place cells, head-direction cells, and grid cells, which together support a map-like representation of space.
  • Path integration lets an animal keep a running estimate of its position by summing self-motion, providing a home vector without external landmarks.
  • Navigational ability varies widely between people and draws on distinct, interacting brain networks rather than a single faculty.

Types of Spatial Orientation

In the Medical Subject Headings vocabulary, Orientation, Spatial sits beneath two broader descriptors that mark its lineage: Orientation — the general awareness of oneself in relation to time, place, and person — and Spatial Behavior, the class of behaviours organised with respect to space. This placement is a classification decision made for indexing the biomedical literature, not a claim that spatial orientation reduces to either parent, and the categories are largely orthogonal to the psychological distinctions drawn below: the egocentric–allocentric divide and the cell-type taxonomy cut across the MeSH tree rather than nesting inside it.

MeSH records one narrower descriptor directly beneath Orientation, Spatial:

- Taxis Response — an innate orienting movement of a whole organism directed toward or away from a stimulus source, such as light or a chemical gradient; the simplest form of spatial orientation, found across the animal kingdom and requiring no internal map.

Taxis Response is listed here as MeSH classifies it, in plain text rather than as a link, because the site does not yet publish a dedicated article on it. The list reflects the structure of the MeSH tree, which is an indexing hierarchy: the presence or absence of a narrower descriptor tracks how the literature is catalogued and does not exhaust the ways spatial orientation can be divided — the reference-frame and neural-code distinctions that organise this article divide the phenomenon along different lines.

Frames of Reference: Egocentric and Allocentric

The first organising distinction in spatial cognition concerns the origin of the coordinate system in which a location is specified. An egocentric representation places the observer at the centre: a cup is to my left, a door is behind me, a turn is to my right. Egocentric coordinates are immediate and directly usable for action — a hand reaches to a body-relative location — but they are fragile, because every one of them changes the instant the body moves. An allocentric representation, by contrast, fixes locations relative to the external world: the cup is on the north edge of the table, the church is east of the square. Allocentric coordinates are stable across the observer's movements and support flexible route planning, but they must be translated back into egocentric terms before the body can act on them (Burgess, 2006).

Neither frame is primary; competent spatial behaviour requires both and a means of converting between them. Reaching and grasping are naturally egocentric; recognising that two routes lead to the same place, or taking a novel shortcut, requires the allocentric map that Tolman first proposed on behavioural grounds (Tolman, 1948). The translation itself is a substantial computation — a rotation and translation of coordinates keyed to the observer's current position and heading — and the brain appears to maintain parallel egocentric and allocentric codes and to transform between them as the task demands, with parietal cortex weighted toward the egocentric and the hippocampal formation toward the allocentric (Ekstrom et al., 2017). The first demonstration makes the distinction concrete: a reader moves and turns a viewer within a fixed array of landmarks and watches the egocentric description of each landmark change while its allocentric description stays put.

Two frames on one scene: egocentric versus allocentric

A landmark (gold) sits due north of the observer. Rotate the observer’s heading and watch the two bearings diverge: the allocentric bearing, measured against north, never moves, while the egocentric bearing, measured against the body’s facing, changes with every turn.

Observer and a fixed landmark in world and body framesThe observer faces a bearing of 40.0°. The landmark sits due north, so its allocentric bearing stays 0 degrees, while its egocentric bearing is 320.0° relative to the observer's facing.NESWlandmark
allocentric axis (north) egocentric axis (facing) landmark

Allocentric bearing to the landmark: 0.0° (fixed). Egocentric bearing: 320.0° — the landmark is 40° to the left. Turning the body rewrites the egocentric bearing while the world-frame bearing holds.

The egocentric bearing is the allocentric bearing minus the heading. Only the body-frame value must be recomputed as the observer moves — the cost the allocentric frame pays once to buy stability.

The Cognitive Map and Its Neural Code

The idea that animals navigate using an internal map, rather than a chain of stimulus–response associations, was argued by Edward Tolman from experiments in which rats took shortcuts and detours that rote learning could not explain; he named the inferred representation a cognitive map (Tolman, 1948). For two decades the cognitive map remained a purely behavioural construct, until recordings from the freely moving rat began to reveal its physical basis. O'Keefe and Dostrovsky found hippocampal neurons — place cells — that fire only when the animal occupies a particular location in an environment, so that the active population signals where the animal is (O'Keefe & Dostrovsky, 1971). A second population, the head-direction cells of the postsubiculum and connected structures, fires as a function of the direction the animal is facing, independent of its location, providing an internal compass (Taube et al., 1990). O'Keefe and Nadel drew these strands together into an explicit theory, arguing that the hippocampus implements an allocentric cognitive map that supports flexible, map-like spatial memory rather than a store of fixed routes (O'Keefe & Nadel, 1978); it is the theoretical bridge from Tolman's behavioural inference to the modern neurobiology.

The map gained a metric with the discovery of grid cells in the entorhinal cortex, neurons that fire at the vertices of a regular triangular lattice tiling the whole environment, as though the brain had laid down a coordinate grid against which distance and direction can be measured (Hafting et al., 2005). Together, place, head-direction, and grid cells — with border and boundary-vector cells that anchor the map to environmental edges — constitute a compact machinery for representing self-location, and the discovery of place and grid cells was recognised with the 2014 Nobel Prize. Figure 1 contrasts the firing signatures of these four cell types. The same architecture is not confined to rodents: functional imaging revealed a grid-like signal in the human entorhinal cortex during virtual navigation, evidence that people carry a homologous mapping system (Doeller et al., 2010). A parallel line of human work identified a cortical region, the parahippocampal place area, that responds selectively to scenes and spatial layout and helps anchor the observer to the surrounding environment (Epstein & Kanwisher, 1998). The history of how these codes were pieced together is itself instructive about how a psychological construct became a neurobiology (Moser et al., 2017).

Figure 1

Firing Signatures of the Four Spatial Cell Types

Four panels showing the characteristic firing patterns of place, grid, head-direction, and boundary cells A place cell fires in one compact region of an enclosure. A grid cell fires at the vertices of a regular triangular lattice covering the whole enclosure. A head-direction cell fires only when the animal faces a narrow range of directions, shown as a tuning wedge on a compass. A boundary cell fires in a band running alongside one wall of the enclosure. Place cell Grid cell Head-direction cell N Boundary cell
Note. Each panel is a schematic firing map. Green, amber, and red shading marks where a cell of that type is active as the animal moves through the enclosure; the compass wedge marks the facing directions that drive the head-direction cell. Schematic; positions are illustrative, not measured data. Original schematic.

Mental Rotation and Spatial Transformations

Keeping oriented also requires transforming spatial representations — imagining how a layout would look from another vantage, or whether a rotated object matches a stored one. The classic demonstration is mental rotation: Shepard and Metzler showed observers pairs of three-dimensional block figures and asked whether they were the same shape or mirror images, and found that the time to decide rose linearly with the angular difference between the two figures, as though the observer were rotating a mental image at a constant angular velocity until the two aligned (Shepard & Metzler, 1971). The linear relation between angle and response time is one of the most reliable results in cognitive psychology, and it made a strong case that at least some spatial thought proceeds through analog, quasi-perceptual transformations rather than discrete propositions.

Mental rotation of an object is closely related to, but distinguishable from, the perspective transformations involved in imagining a scene from a different viewpoint, which recruit the same reference-frame translation that underlies allocentric-to-egocentric conversion. Table 1 contrasts the egocentric and allocentric frames across the dimensions that matter for orientation, and the second demonstration lets a reader run a mental-rotation trial, setting the rotation angle of a figure and observing how the modelled decision time tracks it.

Table 1. The egocentric and allocentric frames of reference contrasted across the dimensions on which they diverge.
Dimension Egocentric frame Allocentric frame
Origin of coordinates The observer's body, eyes, or hand. A landmark or the environment as a whole.
Stability under movement Changes with every move; must be updated constantly. Invariant across the observer's movements.
Directly usable for action Yes; the body acts in body-relative coordinates. No; must be converted to egocentric terms first.
Supports novel shortcuts Poorly; tied to experienced views. Well; supports flexible route planning.
Principal neural weighting Parietal cortex. Hippocampal formation.

Mental rotation: decision time tracks the angle

The reference figure (navy) and a test figure (gold) are the same shape seen at different orientations. Set the angular disparity between them; the model returns the decision time a viewer would take to judge them identical, following the linear angle–time law Shepard and Metzler reported.

Two figures at an angular disparity, with modelled decision timeThe test figure is rotated 120 degrees from the reference. The model returns a decision time of 2500 milliseconds, rising linearly with the angle.referencetest (rotated 120°)decision time2500 ms

At a disparity of 120° the modelled decision time is 2500 ms. Every extra degree adds about 16.7 ms, the signature of an analog rotation running at roughly 60° per second.

Modelled as RT = 500 ms + (1000/60) ms × angle, an illustrative rendering of the classic linear law rather than a fit to any one dataset. Disparity is capped at 180°, the largest a shortest-path rotation requires.

Wayfinding, Path Integration, and Individual Differences

Finding the way through a large environment — wayfinding — draws on the map and its codes but adds a further computation. Path integration is the continuous updating of a position estimate by summing self-motion signals — the distance and direction of each step or turn — so that an animal can compute a straight home vector without ever having seen a landmark. It is the dead-reckoning of the nervous system, and the grid-cell code appears well suited to supplying the metric it needs. Path integration and landmark-based navigation ordinarily work together, the map correcting the drift that self-motion estimates inevitably accumulate.

People differ markedly in how well they orient and navigate, more so than for many cognitive abilities, and those differences are stable and consequential. Navigational skill depends on the quality of the underlying spatial representations, the strategies a person adopts — some rely on an allocentric survey map, others on egocentric route knowledge — and on the integrity of the brain networks that support them (Wolbers & Hegarty, 2010). The dependence on the neural substrate is shown vividly by the finding that London taxi drivers, who train for years on the city's dense street layout, have enlarged posterior hippocampi that grow with experience, a structural correlate of expert spatial memory (Maguire et al., 2000). Navigation is not a single faculty but the coordinated output of interacting systems, and its individual differences reflect variation across the whole ensemble (Ekstrom et al., 2017). The third demonstration builds a path leg by leg and shows the home vector — the direction and distance back to the start — accumulating as path integration would compute it.

Path integration: reading off the home vector

Add legs to the outbound path and watch the home vector — the straight line back to the start — recompute after each step, the way dead reckoning sums self-motion into a running estimate of where home lies. The starting path reproduces the worked example.

Outbound path and the computed home vectorAfter 2 legs, the walker is 50.0 metres from the start. The home vector points at a bearing of 216.9°.startyou
outbound path home vector

2 legs walked. The walker is 50.0 m from the start; the home vector bears 216.9° (south-west). This is the worked example: a 3–4–5 triangle giving a 50 m home vector at 216.9°.

East and north components are summed exactly (heading × distance), then the home vector is their resultant reversed. Legs are fixed values, so the computation is deterministic; the path is capped at six legs.

Worked Example

Consider the core computation of path integration: recovering the straight-line route home from a record of the outbound journey. Suppose a walker leaves home and travels two legs on flat ground. The first leg is 30 metres due east; the second is 40 metres due north. No landmarks are visible, so the return route must be computed from self-motion alone.

Resolve the journey into east and north components. The net eastward displacement is 30 metres and the net northward displacement is 40 metres, so the walker is now at the point (30, 40) relative to home. The straight-line distance from home is the length of that displacement vector, √(30² + 40²) = √(900 + 1600) = √2500 = 50 metres. This is a 3–4–5 right triangle scaled by ten, so the arithmetic is exact.

The direction is found from the same components. The outbound displacement points at a bearing of arctan(east ÷ north) = arctan(30 ÷ 40) = arctan(0.75) = 36.87° east of north — that is, north-east. To walk home, the walker must head in exactly the opposite direction, 36.87° + 180° = 216.87°, which is roughly south-west. The home vector is therefore 50 metres at a bearing of 216.87°.

The point of the example is that this vector was never travelled and never seen: the walker covered 30 + 40 = 70 metres on the way out, yet path integration yields a 50-metre return along a bearing that no single leg followed. Producing that shortcut is exactly the behaviour Tolman took as evidence for a cognitive map (Tolman, 1948), and computing it continuously is what the grid-cell metric is thought to make possible (Hafting et al., 2005). Real path integration also accumulates error with each leg, which is why landmark information is used to reset the estimate whenever it is available.

Discussion

Spatial orientation is one of the clearest cases in which a psychological construct, proposed on purely behavioural grounds, was later given a detailed mechanistic account. Tolman's cognitive map was an inference from the flexibility of animal navigation; the discovery of place, head-direction, and grid cells turned that inference into a description of how the map is physically realised (Tolman, 1948; O'Keefe & Dostrovsky, 1971; Hafting et al., 2005). The two questions that opened this article — the frame of reference and the neural code — turn out to be linked: the allocentric map is built largely in the hippocampal formation, the egocentric representations that drive action are weighted toward parietal cortex, and orientation depends on the traffic between them (Burgess, 2006; Ekstrom et al., 2017).

The account also carries a lesson about levels of explanation. The linear signature of mental rotation and the firing field of a grid cell are findings at different levels — one behavioural, one cellular — yet both describe the same enterprise of representing and transforming spatial relations (Shepard & Metzler, 1971). The field's current direction extends the same machinery beyond physical space, asking whether the map that evolved for navigation is reused to organise knowledge of any kind (Epstein et al., 2017). Spatial orientation, on this view, is not merely one cognitive ability among many but a candidate template for how the brain structures information in general.

Current Directions

The most active line of current work asks whether the brain's spatial machinery is a general-purpose engine for structuring knowledge, not a system dedicated to physical navigation alone. Human imaging of grid-like codes during non-spatial tasks has prompted the proposal that place and grid cells map abstract conceptual spaces — dimensions such as size or meaning — with the same geometry they use for rooms and cities, so that reasoning about concepts may borrow the format of a cognitive map (Bellmund et al., 2018). A closely related programme formalises what a cognitive map actually is in computational terms, treating it as a learned model of the relational structure of a task that supports inference and generalisation, and asking how a single circuit could build such models across domains (Behrens et al., 2018). A third strand revisits the geometry of human spatial memory itself, arguing that people represent large or fragmented environments not with a single Euclidean map but with a mixture of map-like and graph-like structures, stitched together where direct experience connects them (Peer et al., 2021). Running through all three is a synthesis of the animal-physiology and human-cognition traditions into one account of how the mind represents structured relations (Epstein et al., 2017). The trajectory is from mapping physical space toward a general theory of how the brain represents the structure of the world.

Common Misconceptions

A cognitive map is a literal picture of the environment in the head.
The map is a functional representation of spatial relations, not an image. It is realised in the firing of populations of neurons coding location, direction, and distance, and it can be distorted, incomplete, or graph-like rather than a faithful scale drawing (Peer et al., 2021).
Orientation is just perception — knowing where things are.
Orientation is an active, continuous computation. Because every egocentric coordinate changes as the body moves, the system must constantly update self-location and heading, integrating self-motion with sensory input rather than reading off a static scene (Burgess, 2006).
Place and grid cells are unique to rodents.
Although discovered in the rat, the same cell types and a grid-like entorhinal signal have been found in humans, along with a scene-selective cortex that anchors the observer to the surroundings — evidence of a broadly conserved mapping system (Doeller et al., 2010; Epstein & Kanwisher, 1998).
Some people simply have no sense of direction, and nothing can change it.
Navigational ability does vary widely between people, but it rests on strategies and trainable spatial representations, and the hippocampal changes seen in trained navigators show the underlying substrate is plastic rather than fixed (Wolbers & Hegarty, 2010; Maguire et al., 2000).

Glossary

Allocentric reference frame.
A coordinate system that specifies locations relative to the external world or a landmark, invariant across the observer's movements.
Boundary cell.
A neuron that fires when the animal is at a particular distance and direction from an environmental boundary, anchoring the cognitive map to the edges of a space.
Cognitive map.
An internal representation of the spatial relations among places that supports flexible navigation, including novel shortcuts and detours; proposed by Tolman.
Egocentric reference frame.
A coordinate system centred on the observer's body, eyes, or hand, in which locations are specified as directions and distances from the self.
Entorhinal cortex.
A cortical region adjacent to the hippocampus that contains grid cells and supplies the hippocampus with much of its spatial input.
Grid cell.
An entorhinal neuron that fires at the vertices of a regular triangular lattice covering the environment, providing a metric for distance and direction.
Head-direction cell.
A neuron that fires as a function of the direction the animal's head is pointing, independent of location, acting as an internal compass.
Hippocampus.
A medial temporal-lobe structure containing place cells and central to allocentric spatial memory and the cognitive map.
Landmark.
A distinctive, stable environmental feature used to fix position and heading and to correct the drift of path integration.
Mental rotation.
The imagined turning of a spatial figure, whose duration rises linearly with the angle of rotation, taken as evidence for analog spatial transformation.
Parahippocampal place area.
A scene-selective region of human cortex that responds to spatial layout and environmental context, helping anchor the observer to the surroundings.
Path integration.
The continuous updating of a position estimate by summing self-motion signals, yielding a home vector without reference to landmarks.
Place cell.
A hippocampal neuron that fires selectively when the animal occupies a particular location, so that the active population signals current position.
Spatial orientation.
The capacity to represent and continuously update one's position, heading, and the layout of the surrounding environment while moving.
Spatial updating.
The revision of egocentric spatial representations to keep them accurate as the observer moves, so that remembered locations stay correctly placed.
Taxis.
An innate, directed orienting movement of a whole organism toward or away from a stimulus source, the simplest form of spatial orientation.
Wayfinding.
The purposeful process of determining and following a route between an origin and a destination through a large-scale environment.

Key Researchers

Neil Burgess (contemporary). Cognitive and computational neuroscientist at University College London; builds computational models of place, grid, and boundary cells and clarified how egocentric and allocentric spatial codes combine, and co-reported grid-like signals in the human brain. ORCID - Faculty Page - Wikipedia

Russell A. Epstein (contemporary). Cognitive neuroscientist at the University of Pennsylvania; co-discovered the parahippocampal place area and studies how scene perception and cortical spatial codes support human orientation and navigation. ORCID - Faculty Page - Wikipedia

Eleanor A. Maguire (1970-2025). Cognitive neuroscientist at University College London; her studies of London taxi drivers showed that intensive navigational experience is associated with enlarged posterior hippocampi, a landmark demonstration of experience-dependent plasticity in spatial memory. ORCID - Wikipedia - Wikidata

Edvard I. Moser (contemporary). Neuroscientist at the Kavli Institute for Systems Neuroscience, NTNU; co-discovered grid cells in the entorhinal cortex and shared the 2014 Nobel Prize for uncovering the brain's positioning system. ORCID - Faculty Page - Wikipedia

May-Britt Moser (contemporary). Neuroscientist and co-director of the Kavli Institute for Systems Neuroscience, NTNU; co-discovered grid cells and shared the 2014 Nobel Prize for work on the neural representation of space. ORCID - Faculty Page - Wikipedia

Lynn Nadel (contemporary). Cognitive psychologist at the University of Arizona; co-author with John O'Keefe of The Hippocampus as a Cognitive Map, which set out the theory that the hippocampus implements an allocentric spatial map. Faculty Page - Wikipedia - Wikidata

John O'Keefe (contemporary). Neuroscientist at University College London and director of the Sainsbury Wellcome Centre; discovered place cells in the hippocampus and shared the 2014 Nobel Prize for the discovery of the brain's cognitive map. ORCID - Faculty Page - Wikipedia

Edward C. Tolman (1886-1959). Psychologist at the University of California, Berkeley; introduced the concept of the cognitive map from studies of maze learning in rats, arguing that animals acquire flexible spatial knowledge rather than fixed responses. Wikipedia - Wikidata

Frequently Asked Questions

What is spatial orientation?
It is the ability to represent one's position, heading, and the layout of the environment, and to keep that representation current as the body moves; it is an active computation rather than a static perception (Burgess, 2006).

What is the difference between egocentric and allocentric frames of reference?
Egocentric coordinates are centred on the observer and change with every movement, while allocentric coordinates are fixed to the world and stay constant; action uses egocentric coordinates, but flexible route planning uses the allocentric map (Burgess, 2006).

What is a cognitive map?
It is an internal representation of the spatial relations among places that lets an animal take shortcuts and detours; Tolman inferred it from rat behaviour, and later work identified the neurons that realise it (Tolman, 1948; O'Keefe & Dostrovsky, 1971).

What are place cells, grid cells, and head-direction cells?
Place cells fire when an animal is in a particular location, head-direction cells fire according to which way it faces, and grid cells fire on a regular lattice that supplies a metric for distance; together they form a neural positioning system (O'Keefe & Dostrovsky, 1971; Taube et al., 1990; Hafting et al., 2005).

What is path integration?
It is the continuous updating of a position estimate by summing self-motion signals, so that an animal can compute a straight home vector without seeing any landmark; the estimate accumulates error and is reset by landmarks when available (Hafting et al., 2005).

Why does mental rotation take longer for larger angles?
Because observers appear to rotate a mental image at a roughly constant rate until it aligns with the target, so decision time rises linearly with the angular difference, which is evidence for analog spatial transformation (Shepard & Metzler, 1971).

Do humans have the same spatial cells as rats?
Yes; a grid-like signal has been recorded in the human entorhinal cortex, and humans also have a scene-selective cortical area that anchors orientation, indicating a broadly conserved mapping system (Doeller et al., 2010; Epstein & Kanwisher, 1998).

Why are some people better at navigating than others?
Navigational ability varies widely and reflects differences in spatial strategies, the quality of internal representations, and the brain networks that support them; the substrate is plastic, as the enlarged hippocampi of trained taxi drivers show (Wolbers & Hegarty, 2010; Maguire et al., 2000).

References

Behrens, T. E. J., Muller, T. H., Whittington, J. C. R., Mark, S., Baram, A. B., Stachenfeld, K. L., & Kurth-Nelson, Z. (2018). What is a cognitive map? Organizing knowledge for flexible behavior. Neuron, 100(2), 490-509. https://doi.org/10.1016/j.neuron.2018.10.002

Bellmund, J. L. S., Gardenfors, P., Moser, E. I., & Doeller, C. F. (2018). Navigating cognition: Spatial codes for human thinking. Science, 362(6415), eaat6766. https://doi.org/10.1126/science.aat6766

Burgess, N. (2006). Spatial memory: How egocentric and allocentric combine. Trends in Cognitive Sciences, 10(12), 551-557. https://doi.org/10.1016/j.tics.2006.10.005

Doeller, C. F., Barry, C., & Burgess, N. (2010). Evidence for grid cells in a human memory network. Nature, 463(7281), 657-661. https://doi.org/10.1038/nature08704

Ekstrom, A. D., Huffman, D. J., & Starrett, M. (2017). Interacting networks of brain regions underlie human spatial navigation: A review and novel synthesis of the literature. Journal of Neurophysiology, 118(6), 3328-3344. https://doi.org/10.1152/jn.00531.2017

Epstein, R., & Kanwisher, N. (1998). A cortical representation of the local visual environment. Nature, 392(6676), 598-601. https://doi.org/10.1038/33402

Epstein, R. A., Patai, E. Z., Julian, J. B., & Spiers, H. J. (2017). The cognitive map in humans: Spatial navigation and beyond. Nature Neuroscience, 20(11), 1504-1513. https://doi.org/10.1038/nn.4656

Hafting, T., Fyhn, M., Molden, S., Moser, M.-B., & Moser, E. I. (2005). Microstructure of a spatial map in the entorhinal cortex. Nature, 436(7052), 801-806. https://doi.org/10.1038/nature03721

Maguire, E. A., Gadian, D. G., Johnsrude, I. S., Good, C. D., Ashburner, J., Frackowiak, R. S. J., & Frith, C. D. (2000). Navigation-related structural change in the hippocampi of taxi drivers. Proceedings of the National Academy of Sciences, 97(8), 4398-4403. https://doi.org/10.1073/pnas.070039597

Moser, E. I., Moser, M.-B., & McNaughton, B. L. (2017). Spatial representation in the hippocampal formation: A history. Nature Neuroscience, 20(11), 1448-1464. https://doi.org/10.1038/nn.4653

O'Keefe, J., & Dostrovsky, J. (1971). The hippocampus as a spatial map. Preliminary evidence from unit activity in the freely-moving rat. Brain Research, 34(1), 171-175. https://doi.org/10.1016/0006-8993(71)90358-1

O'Keefe, J., & Nadel, L. (1978). The hippocampus as a cognitive map. Oxford University Press. ISBN 9780198572060.

Peer, M., Brunec, I. K., Newcombe, N. S., & Epstein, R. A. (2021). Structuring knowledge with cognitive maps and cognitive graphs. Trends in Cognitive Sciences, 25(1), 37-54. https://doi.org/10.1016/j.tics.2020.10.004

Shepard, R. N., & Metzler, J. (1971). Mental rotation of three-dimensional objects. Science, 171(3972), 701-703. https://doi.org/10.1126/science.171.3972.701

Taube, J. S., Muller, R. U., & Ranck, J. B. (1990). Head-direction cells recorded from the postsubiculum in freely moving rats. I. Description and quantitative analysis. Journal of Neuroscience, 10(2), 420-435. https://doi.org/10.1523/JNEUROSCI.10-02-00420.1990

Tolman, E. C. (1948). Cognitive maps in rats and men. Psychological Review, 55(4), 189-208. https://doi.org/10.1037/h0061626

Wolbers, T., & Hegarty, M. (2010). What determines our navigational abilities? Trends in Cognitive Sciences, 14(3), 138-146. https://doi.org/10.1016/j.tics.2010.01.001