45 Drawing the Map
How Movement, Landmarks, and Experience Become Neural Structure
45.1 Begin with the animal
An animal leaves a refuge, turns around obstacles, searches for food, and eventually must return. The goal may be hidden. The outward route may not be available in reverse. No single smell, landmark, or familiar turn will always provide the answer. Successful navigation requires the animal to relate its own movement to a world that remains only partly visible.
Chapter 44 established the comparative evidence for an ancient hippocampal or medial-pallial system. The question here is how movement and sensation produce usable neural relations. How can activity keep track of position, distinguish two journeys through the same place, and preserve the order in which one location leads to another?
Navigation is the biological starting point of this explanation because its demands are concrete. The animal needs to know which way it is facing, how far it has traveled, where boundaries lie, and which routes connect its present location to food, water, shelter, or escape. These relations matter before there is language or a narrated personal history. That explanatory priority does not establish that every nonspatial relational capacity evolved later; the comparative evidence considered in the preceding chapter does not support such a simple chronology.
Most of the detailed physiology that follows comes from rats and mice. It reveals operations of a living mammalian circuit, not a direct recording of an ancestral vertebrate brain. Within that scope, it allows us to trace a useful progression: self-motion updates an estimate, environmental features anchor it, experience differentiates situations, and coordinated neural activity expresses transitions extending beyond the body’s present position.
The map takes shape in the relation between a moving body and a structured world. Its usefulness depends on preserving those relations well enough to guide the next action.
45.2 A place enters the physiology
The behavioral argument introduced earlier can now be given a physiological form. Tolman’s cognitive map was an inference from flexible performance: behavior sometimes reflected relations among places that could not be explained by one reinforced chain of turns [@tolman1948cognitive]. What would such relations look like in neural activity?
John O’Keefe and Jonathan Dostrovsky supplied the first direct physiological evidence. In 1971 they recorded single neurons from the hippocampus while rats moved freely through an enclosure. Some neurons fired little through most of the environment but became active when the rat occupied a restricted region. O’Keefe called these neurons place cells, and the region associated with elevated firing became the cell’s place field [@okeefe1971spatial].
A place field is not a tiny picture or a mark printed on an internal chart. Investigators construct a firing-rate map by dividing the enclosure into spatial bins, measuring how long the animal occupied each bin, and calculating the cell’s firing rate there. The resulting colored patch shows a statistical relation between neural activity and the animal’s position. It is a view of the data, not an image found inside the rat.
The physiological result is still remarkable. During navigation, activity across a population of hippocampal neurons can distinguish one part of a familiar environment from another. A decoder fitted to the activity of many cells can estimate where the animal is. This demonstrates information available in the recorded population; it does not by itself identify how another neural population reads that information. The map is a relation among population activity, the moving body, and the environment.
Place cells also make a claim about what the hippocampus does not require. A location does not have one defining stimulus. The animal can occupy the same place while facing different directions, receiving different retinal images, or sampling different odors. Place-related activity can persist in darkness for a time and can remain recognizable as particular cues change. The system integrates multiple sources of evidence rather than waiting for one sensory feature to announce the answer.
The simplest description is useful: place cells tell the system where the animal is. It becomes misleading only when treated as complete. A cell can have more than one field, fields are not distributed as a uniform set of pixels, and the active population can change when the behavioral meaning of an environment changes. The hippocampus represents a place as part of a situation, not as an isolated coordinate.
45.3 The same location can be a different situation
In Loren Frank and colleagues’ experiment, rats traveled among the center and outer arms of a W-shaped track. Traversals could share a segment of track while belonging to different journeys. Hippocampal and entorhinal neurons sometimes fired differently on that shared segment according to the preceding or upcoming arm. The animal’s paws could occupy the same floorboards while its population activity distinguished the route [@frank2000trajectory].
This result changes the interpretation of a place field. Physical position remains important, but position is not the whole represented state. Route, direction, recent history, expected destination, and task demands can alter the population active at that position. A useful navigational system must make those distinctions. The same junction can require a left turn when returning to shelter and a right turn when approaching food.
The population can also reorganize when the larger context changes. In rate remapping, many cells retain fields in roughly the same locations but change how strongly they fire. In global remapping, the spatial relationships among the cells’ fields reorganize substantially; cells may acquire, lose, or relocate fields. Fyhn and colleagues found entorhinal grid realignment accompanying hippocampal global remapping, whereas rate remapping could occur with stable grid fields [@fyhn2007remapping]. These are distinguishable population changes, not simply two names for a cell firing differently. Route-dependent firing at a shared location need not imply global remapping of the environment.
Remapping shows why stability cannot be the only criterion for a useful spatial code. An animal needs enough stability to recognize a place across ordinary changes in viewpoint, illumination, and sensory noise. It also needs enough flexibility to distinguish two situations that share the same geometry but predict different outcomes. A stable map that ignored context would repeatedly send the animal toward the wrong goal.
The map is therefore not a photograph of the surroundings. It is an organized estimate of the animal’s relation to a world in which particular actions lead to particular consequences. Spatial location supplies a scaffold. Route, context, and goal can change what that location means for the next action. This is a functional interpretation of the population distinctions; demonstrating a distinction does not establish which downstream decision it controls.
45.4 A circuit, not a row of instruments
The recurrent organization introduced in Chapter 43 matters directly to the physiology. Place cells participate in a system of hippocampal, entorhinal, and adjoining populations that influence one another. There is no requirement that one region first calculate a complete coordinate for another region to label. The response classes discovered within this system identify useful relations, but they are not independent instruments assembled into a navigation console.
The discovery of grid cells made that temptation especially strong. When a rat forages across an open floor, many neurons in medial entorhinal cortex fire in several locations. Plot those locations and they often approximate a triangular lattice, producing sixfold or hexagonal symmetry [@hafting2005microstructure]. Different cells have different spatial phases, and groups of cells form modules with related spacing and orientation. Grid spacing tends to increase from dorsal toward more ventral entorhinal levels [@stensola2012modules].
This pattern is well suited to representing displacement. As the animal moves, the population changes in a regular way. Multiple modules operating at different spatial scales can jointly distinguish positions over a range much larger than the spacing of any single module. Grid activity has therefore become central to models of path integration, continuous attractor dynamics, and multiscale position coding.
The grid is not an invariant sheet of graph paper laid over every environment. Change the enclosure geometry and the pattern can distort, reorient, or become less regular. Environmental geometry shapes grid symmetry [@krupic2015geometry]. Peng and colleagues made this dependence especially clear in a homing task. Mice left a home base, searched an arena for a movable lever, pressed it, and returned home for food. During dark trials, grid-cell firing retained movement-related structure even though maps accumulated across trials lacked a stable lattice. Activity was anchored differently during search and homing, including reanchoring to the lever after the encounter. Drift in the population’s represented movement direction predicted the animal’s homing direction [@peng2025referenceframes]. A poor room-centered firing map therefore need not mean that the population has ceased to track movement. The recording links the neural estimate to behavior without establishing that grid cells alone generate the homing command.
Grid activity should therefore be described as an internally organized, movement-sensitive population code that contributes to estimating position. It is not a complete Cartesian coordinate system independent of landmarks, task, and hippocampal feedback.
That last point matters anatomically. Medial entorhinal lesions reduce the precision and long-term stability of hippocampal place fields, but they do not erase every place field. Influence also runs in the other direction. Inactivating the hippocampus disrupts the periodic organization of entorhinal grid firing [@hales2014mec; @bonnevie2013grid]. These perturbations establish functional dependence in both directions. They do not reduce that dependence to a single direct synapse or prove that every place field requires an intact grid pattern.
Other recurring signals contribute to that loop.
Head-direction cells fire according to the direction the animal faces in the horizontal plane. A cell may prefer northeast and remain quiet when the animal faces southwest, regardless of where the animal stands. Populations of these cells provide an allocentric directional signal, often described as a neural compass [@taube1990head].
Border cells fire strongly when the animal is near one or more environmental boundaries [@solstad2008borders]. Boundary-vector cells are defined more specifically by a boundary at a preferred distance and allocentric direction from the animal: for example, a wall a certain distance to the east, regardless of which way the head faces [@lever2009boundary]. The two terms should not be treated as synonyms, although both identify ways in which environmental edges organize activity.
Speed and velocity signals change with linear or angular movement. Some medial entorhinal neurons increase their firing rate with running speed, while other populations carry combinations of position, direction, speed, and theta-related information [@kropff2015speed; @hardcastle2017multiplexed]. These variables are not housed in perfectly separate boxes. Many neurons show mixed selectivity, and the strength of their tuning changes with behavior.
The named cell classes remain useful because each made one relation conspicuous. Head-direction cells exposed direction. Grid cells exposed periodic spatial organization. Border cells exposed environmental edges. Speed cells exposed movement rate. But the animal is not navigated by a row of independent gauges. The map arises from interacting populations that combine body movement with external structure.
The firing fields of many grid cells approximate sixfold symmetry because neighboring fields are arranged at roughly 60-degree intervals. That geometry is an empirical finding. Why the network produces it remains an active mechanistic question.
Several classes of model can generate periodic activity. Continuous-attractor models organize a population so that an activity pattern moves smoothly as the animal moves. Oscillatory models derive spatial periodicity from interference among movement-sensitive rhythms. Efficient-coding and self-organization accounts ask what representations emerge when a network learns from natural trajectories. These models make different assumptions, and no single mechanism has closed the question.
Grid modules add a second observation. Cells within a module share similar spacing and orientation but differ in phase. Modules with different scales can be compared to gears with different numbers of teeth: the joint configuration can distinguish many more positions than any one gear alone. This gives the population enormous coding capacity. It does not prove that downstream neurons read the code as an odometer, nor that grid activity alone determines every place field.
Grid populations provide a highly structured signal about movement and position. Their modular organization offers a plausible basis for multiscale spatial estimation, while their distortions reveal continuing dependence on environmental geometry and task.
45.5 Keeping track while moving
The most fundamental problem is not recognizing a familiar landmark. It is keeping track of position when the landmark is behind the animal, hidden by an obstacle, or temporarily absent.
Many mobile animals solve part of this problem through path integration. The animal begins from a known location and continually updates an estimate of its position from its own movement. Turning changes heading. Movement changes the estimated displacement from the starting point. Integrating those changes can produce a direct homeward vector even after a winding outward path.
Desert ants offer a striking behavioral example. An ant leaves its nest and follows a long, irregular search path. After finding food, it often turns toward the nest and takes a relatively direct route home. The ant does not need to retrace every outbound turn. It has accumulated information about direction and distance while moving [@muellerwehner1988path]. Ants do not possess a vertebrate hippocampus; they isolate the computational advantage of path integration without identifying the vertebrate circuit that performs it.
A vertebrate brain has several sources for the same general calculation. The vestibular system reports rotations and linear acceleration. Proprioception and touch report movement of the limbs and body. Optic flow provides visual evidence of self-motion. Motor-related signals indicate actions that have just been issued. Head-direction and speed-related populations supply variables needed to update the estimate.
Path integration is powerful because it continues to operate when external evidence is poor. It is also inevitably imperfect. Small errors in estimating each turn or segment accumulate. Without correction, uncertainty generally grows with time and distance. Individual errors can partly cancel, but the estimate has no guarantee of remaining aligned with the world.
This limitation reveals why the external world remains part of the computation. Boundaries and landmarks provide correction. In rodent recordings, grid firing drifted with time and distance along long excursions away from a boundary. Encounters with boundaries were associated with reduced error, especially in the spatial direction constrained by the encountered boundary [@hardcastle2015boundaries]. The result concerns drift in the neural firing pattern, not a direct measurement of the animal’s subjective uncertainty. Border-cell input is a plausible correction mechanism supported by modeling, rather than a pathway selectively proved necessary by that experiment. Landmarks can rotate or reanchor head-direction and place representations. Familiar geometry can help keep an internally updated estimate aligned with the environment.
The relation is reciprocal. Self-motion keeps the map moving when landmarks are unavailable. Landmarks and boundaries correct the drift produced by self-motion. Neither source is sufficient by itself across the conditions a freely moving animal encounters.
This interaction also explains why the grid pattern changes with the environment. If the code were an isolated metric, environmental distortion would count only as failure. In a control system, calibration to the world is part of successful operation. A local object can become the relevant reference for homing. A barrier can divide one open surface into routes that are no longer interchangeable. A wall can supply the positional evidence needed to reset accumulated error.
Movement also changes the evidence available for correction. Turning toward a landmark or approaching a boundary can improve the estimate used to guide the next movement. Action and sensing therefore form a loop: the animal moves partly on the basis of its estimate, and the resulting sensory encounter can revise that estimate. This is the embodied character of the computation.
Path integration carries the estimate forward when sensory evidence is sparse. Boundaries and landmarks keep the estimate from drifting away from the world.
45.6 Development assembles and anchors the system
Development helps separate the internal organization of a population from its alignment with the environment.
Before rat pups open their eyes, head-direction cells can already maintain coherent relations with one another. Preferred directions drift relative to the room, but simultaneously recorded cells tend to drift together. After eye opening, directional tuning becomes more stable [@bjerknes2015coherence]. The distinction is important: a network can preserve an internal ordering while its reference to the outside world remains poorly anchored.
These observations show that patterned vision through open eyes is not necessary for the first expression of directional coherence. They do not show that the network developed without sensory input or activity. The pups had vestibular and other nonvisual experience, and maturation continued throughout the recordings. Vision becomes available to calibrate a system that is already organized; the observations do not isolate every cause of its earlier organization.
When young rats first explore beyond the nest, head-direction responses are comparatively mature. Place-related firing is also present, although its precision and stability continue to improve. Regular, stable grid firing develops later [@langston2010development; @wills2010development]. Thus, adult-like grid periodicity cannot be a prerequisite for the earliest place fields. That conclusion is narrower than claiming that early place firing is independent of entorhinal input: immature entorhinal populations can carry useful signals without expressing a mature grid lattice.
Nor does the order of emergence prove that head-direction cells cause place fields, which then cause grid cells. Different responses may depend on shared maturation, movement, sensory calibration, and reciprocal connections. Establishing a particular causal dependency requires perturbing the relevant signal or pathway, not simply observing which response appears first.
A rat’s postnatal development also does not replay vertebrate evolutionary history. These recordings explain how components of one mammalian system become coordinated. They support the chapter’s central argument at that level: internal organization and environmental calibration are distinguishable contributions that become coupled as the animal acts.
45.7 From location to route
A coordinate answers one question: where is the animal? Successful navigation requires another: what can be reached from here?
A wall makes the distinction obvious. Two locations can be physically close but separated by a barrier. A distant doorway may be the only available transition between them. Straight-line distance therefore does not determine the next useful action. The animal must learn how places are connected.
The term state can sound more abstract than the biology requires. In this chapter, a state means the information needed to distinguish the animal’s current navigational situation. It can include position, heading, route, nearby boundaries, recent movement, and the current destination. Two visits to the same coordinate can be different states when they make different transitions possible.
A transition is movement from one such state to another. Repeated travel teaches which transitions are available, which are blocked, which are fast, and which lead toward a goal. A route is an ordered series of transitions. A shortcut becomes possible when the animal can use the learned relations without simply replaying one trained motor sequence.
This is the point at which a spatial map becomes predictive. Current population activity does not need to contain a picture of the future. It needs to be shaped by the states that tend to follow the present state and by the routes through which they are reached.
The successor representation gives one formal account of that idea. A state is represented through the expected, temporally discounted occupancy of states under a particular policy, the rule or tendency governing which actions are taken. More imminent and more frequently expected visits contribute more strongly; proximity here concerns expected transitions, not straight-line distance. Stachenfeld and colleagues used this formalism to explain how spatial firing could reflect barriers, routes, and experience [@stachenfeld2017predictive].
The qualification about policy matters. If an animal repeatedly follows one route, its learned successor representation need not include an unused alternative simply because that alternative is physically possible. A newly encountered barrier changes available transitions, but a cached representation will not necessarily update everywhere at once. Changes require learning, inference, or additional computation. The model makes a testable claim about predictive structure, rather than supplying an all-purpose planner. Hippocampal populations also carry sensory, contextual, temporal, and goal-related information that no single formalism exhausts.
The biological problem remains reachability. A barrier can make a nearby refuge difficult to reach, while a familiar corridor makes a more distant refuge accessible. Learning that structure and valuing a destination are distinguishable operations. A thirsty animal may assign new value to a known water source without relearning the entire environment, yet still need to revise its route if the connecting path has changed. This distinction will matter when particular encounters update a general map in Chapter 47.
45.8 The map moves ahead of the animal
So far the map has been described mainly through firing rates accumulated across movement. Neural activity also has temporal structure within a traversal. An average rate map discards much of that structure.
During active exploration in rats, hippocampal activity is organized by a prominent theta rhythm, commonly about 6 to 12 hertz. As a rat crosses a place field, the corresponding cell often fires at progressively earlier phases of successive theta cycles. O’Keefe and Michael Recce described this relation as phase precession [@okeeferecce1993phase]. It relates the timing of one cell’s spikes to the animal’s progress through its field.
A theta sequence is a population phenomenon. Within one theta cycle, cells with overlapping fields can fire in the order in which those fields lie along a route, compressing a segment of movement into roughly a tenth of a second. A sequence may extend from locations behind the body to locations ahead of it. Foster and Wilson demonstrated such ordered sequences. Shuffling spike timing in the recorded data, while preserving individual phase–position relations, greatly reduced sequence organization [@fosterwilson2007theta]. Single-cell phase precession and coordinated population order therefore require separate measurements.
The circuit provides a way to understand why. CA1 receives both input from CA3 through Schaffer collaterals and direct input from entorhinal layer III onto more distal dendrites. Their timing and relative influence can change the temporal organization of CA1 output. In mice running on a familiar linear track, selective inhibition of layer III terminals in CA1 reduced the slope and phase range of precession while strengthening population theta sequences. Spatial firing was largely preserved [@guardamagna2025entorhinal]. Related recordings distinguish network states with different degrees of phase precession and sequence organization [@guardamagna2023heterogeneity].
This intervention argues against a simple chain in which more phase precession must always produce stronger sequences. It supports a role for interactions between input streams and circuit dynamics, while leaving the complete sequence-generating mechanism unresolved. The recurrent CA3 network is a candidate contributor to ordered activity, and the direct entorhinal input can alter its expression in CA1; those roles should not be turned into an exclusive division between stored past and sensory present.
Temporal compression makes successive locations available to interacting neurons within a short interval. This could help downstream circuits compare a current state with an approaching one, and could support changes in connections between coactive assemblies. The timing observation alone does not identify the plasticity rule or demonstrate that a particular sequence formed a memory. Nor does every cell precess on every traversal, or every theta cycle contain a clean forward sequence.
During movement, hippocampal populations can express an ordered spatial segment extending beyond the body’s instantaneous position. The ordering is a property of coordinated activity, not merely a collection of colored place fields.
45.9 At a choice point
The pauses and head turns rats make at difficult junctions are known as vicarious trial and error. Johnson and Redish recorded hippocampal CA3 ensembles during decision-point behavior and decoded activity representing paths ahead of the animal, sometimes sweeping along one available route and then another [@johnsonredish2007paths]. The body had not traversed those paths during the decoded event.
Such activity is a candidate mechanism for considering alternatives. The measured result is nonlocal, ordered population activity; the stronger claim that the sequence evaluates outcomes or determines the ensuing choice needs additional evidence. The decoder reveals neither a conscious image nor the operation of an inner observer.
Yu and colleagues separated goal direction from heading and movement using recordings from the Honeycomb maze. A rat occupies one platform until two neighboring platforms rise, then chooses between them on its way to a remembered goal. CA1 theta sweeps were biased toward the goal even when heading and movement pointed elsewhere. Stronger goal alignment preceded correct choices. Goal modulation also increased within sessions in already trained animals, without a corresponding reliable improvement in performance across those trials [@yu2026thetasweeps].
The study strengthens the case that theta sweeps can supply information useful for goal-directed action. It does not selectively perturb the goal-directed component to establish that it causes the choice. Its network model offers a possible implementation of goal-biased dynamics, rather than identifying the unique biological source of the goal signal.
The control problem therefore extends beyond the hippocampus. A represented destination acquires significance through interactions with systems concerned with bodily need, outcome value, task rules, and action selection. The navigation recordings do not, by themselves, establish a direct projection from every such system to the recorded cells. They establish that an absent goal can organize hippocampal activity differently from the direction in which the body is currently oriented.
45.10 When the body stops
A different population state appears during quiet wakefulness and non-REM sleep. A sharp-wave ripple combines a brief population event with a fast oscillation, physiologically distinct from movement-related theta. Some of these events contain ordered firing compressed into tens to hundreds of milliseconds. A ripple is detected from the population field signal; demonstrating a sequence requires examining the spikes or decoded activity within it. The two observations are not interchangeable.
Lee and Wilson found that CA1 ensembles could express, during slow-wave sleep after experience, the order of place-cell firing observed during earlier movement [@leewilson2002sequential]. During quiet wakefulness, sequences can also run in reverse, as Foster and Wilson observed after rats traversed a track [@fosterwilson2006reverse]. Other events express trajectories biased from the current location toward a remembered goal, as in Pfeiffer and Foster’s open-arena task [@pfeifferfoster2013future]. The term replay is useful, but it should not imply that every ripple reproduces one recorded episode exactly.
These possibilities have different functional implications. Reactivating an experienced order could reinforce relations learned during travel. A reverse sequence after reward could help connect an outcome with preceding states. A sequence toward a remembered goal could make a route available before movement. Those are mechanistic proposals about what ordered activity can accomplish; the direction of a sequence does not by itself certify its function.
Interventions provide stronger evidence that ripple-associated activity matters under particular conditions. Girardeau and colleagues detected ripples and interrupted them with brief electrical stimulation during post-training rest and sleep. Rats learned a spatial reference-memory task less well than controls receiving delayed stimulation [@girardeau2009ripples]. Jadhav and colleagues disrupted awake ripples during W-track alternation. The deficit was concentrated in outbound choices requiring the animal to distinguish its recent route; inbound returns to the center were spared [@jadhav2012ripples].
These experiments connect ripple-associated events to later spatial performance and to memory-guided choices during ongoing behavior. Their scope matters: interrupting a ripple also disturbs accompanying neuronal discharge. It does not isolate the oscillation from the represented content, prove the necessity of one decoded trajectory, or establish that every ripple consolidates a memory. The sleep and awake interventions address different uses of recently acquired information.
Theta and ripple sequences thus provide complementary means of expressing order. Theta organizes population activity during active engagement with the environment, including sampling and choice. Ripple-associated events can express ordered relations during pauses and sleep, when the changing sensory stream no longer dictates the succession. Neither state has one exclusive psychological function.
45.11 What an embodied map makes possible
The physiology gives substance to the control argument. Self-motion helps carry an estimate through gaps in sensation. Environmental features constrain accumulated error. Population distinctions separate journeys and contexts, while learned transitions preserve how one situation leads to another. Sequences allow some of those relations to become active when the corresponding movement is not being performed.
This is a plausible biological foundation for using experience beyond the moment in which it was acquired. It does not require a human autobiographical system, and it is compatible with the ancient spatial-relational functions considered in Chapter 44. It also leaves room for lineages to elaborate different capabilities. The rodent experiments cannot tell us which temporal and nonspatial relations accompanied navigation in the earliest members of that history.
The same mechanisms create a bridge to the problem of particular encounters. A route can be represented across many journeys, yet a useful memory must sometimes distinguish today’s blocked passage from yesterday’s open one. Route- and context-sensitive populations offer ways to keep overlapping experiences distinguishable; sequential activity offers ways to express their order. Those operations are ingredients for an account of episodic memory, not evidence that a decoded rat trajectory is equivalent to human recollection.
What remains to be explained is how event-specific content modifies a more general structure and is retrieved when the animal’s needs change. That is the question reserved for Chapter 47. Here the physiological contribution is narrower and concrete: a population code can preserve both relations among states and the order in which they become active, while remaining coupled to a body that senses and acts.
45.12 Coda: from places to other spaces
This chapter began with an animal leaving a refuge and ended with population activity extending beyond its present location. Place-related firing identifies locations within context. Grid, direction, boundary, speed, and sensory signals contribute to updating and anchoring the representation. Remapping differentiates situations, and learned transitions distinguish physical proximity from practical reachability. Theta and ripple-associated sequences express relations in an order that need not match what the body is doing now.
Physical navigation provides an unusually clear test because position, movement, boundaries, and routes can be measured against neural activity. The biological benefit is equally concrete: an animal can reach what it needs when the destination is absent from current sensation.
Chapter 46 asks how far these organizing principles extend to tones, concepts, task states, and social relations. The challenge is to establish what the neural organization explains in each domain. Navigation supplies the starting point for that inquiry without deciding in advance when every other relational capacity evolved.
Established findings. In the rodent experiments considered here, hippocampal and entorhinal populations carry information about location, route, direction, boundaries, movement, and context. Spatial activity reflects both self-motion and environmental anchoring. During theta states and some sharp-wave ripples, ensembles express ordered relations beyond the animal’s current position. Perturbations establish contributions of particular inputs and ripple-associated events under specified conditions.
The chapter’s organizing interpretation. These interacting operations make sense as part of embodied control. An animal benefits from preserving relations that remain useful when a goal or landmark is absent, and from revising those relations when its actions reveal a changed world. Navigation makes that ancient biological problem visible; it does not establish a universal chronology from spatial to nonspatial cognition.
Open questions. No single mechanism accounts for every grid pattern or every hippocampal sequence. Developmental timing does not fully identify causal dependencies. Decoded trajectories and goal biases do not, by themselves, show which activity is used to select an action, and ripple interruption does not isolate one sequence’s mnemonic function. How these operations preserve the specificity of an encounter remains a further question.