Cognitive Map Hippocampal Theory – John O’Keefe & Lynn Nadel

The quest to understand how sentient organisms orient, navigate, and preserve a coherent sense of reality within a dynamic physical world represents one of the most profound inquiries at the intersection of psychology, neurobiology, and philosophy. For centuries, the nature of space was the exclusive domain of philosophers and mathematicians, who debated whether our comprehension of the external environment is an empirical construct derived piecemeal from sensory impressions or an innate architectural framework imposed by the mind upon sensory chaos. In the mid-twentieth century, this conceptual inquiry migrated into the physiological laboratory, catalyzed by breakthroughs in behavioral psychology and microelectrode engineering, transforming abstract epistemological queries into empirical neurobiology.

At the center of this paradigm shift was the formulation of the Cognitive Map Theory of hippocampal function, articulated by neurophysiologist John O’Keefe and cognitive psychologist Lynn Nadel in their landmark 1978 monograph, The Hippocampus as a Cognitive Map. By synthesizing cutting-edge in vivo single-unit electrophysiology with rigorous ethological analysis and philosophical epistemology, O’Keefe and Nadel proposed a revolutionary concept: that the mammalian hippocampus does not merely process generic associative learning or general memory consolidation, but rather serves as an internal, allocentric representation of absolute physical space. This neural metric framework—instantiated biologically by the spatially selective firing of pyramidal neurons termed “place cells”—freed animals from the rigid constraints of stimulus-response contingencies, providing an internal coordinate system capable of computing novel trajectories, estimating spatial geometry, and supporting flexible navigation.

The implications of O’Keefe and Nadel’s thesis extended far beyond animal spatial orientation. By conceptualizing the hippocampus as an active coordinate matrix, their framework resolved long-standing clinical paradoxes regarding the profound memory deficits seen in human amnesic patients such as H.M., while providing an evolutionary scaffolding upon which complex episodic memory, imaginative foresight, and symbolic conceptual navigation could emerge. The cognitive map theory transformed spatial neurobiology from a fringe subdiscipline into a cornerstone of contemporary cognitive neuroscience, bridging cellular physiology, network dynamics, computational modeling, and neuropsychology. The following exploration details the origins, empirical validation, anatomical substrates, and enduring legacy of this revolutionary framework.

1. Historical Foundations and the Genesis of Spatial Cognition

1.1 Edward Tolman and the Concept of Cognitive Maps

The intellectual ancestry of spatial cognition resides in the radical theoretical work of Edward C. Tolman. During the 1930s and 1940s, American experimental psychology was dominated by strict behaviorist paradigms, spearheaded by figures such as John B. Watson and Clark Hull. Within this orthodox behaviorist framework, all animal behavior was conceptualized as an unbroken chain of peripheral reflexes: associations stamped into the nervous system through blind stimulus-response (S-R) pairings governed strictly by instantaneous drive reduction and reinforcement. Organisms were viewed as passive automatons navigating environments via mechanical habit chains, executing fixed sequences of muscle contractions (egocentric motor responses) triggered by proximate sensory stimuli.

Tolman vehemently challenged this peripheralist dogma. In his seminal 1948 treatise, “Cognitive Maps in Rats and Men”, Tolman asserted that animals build complex, holistic internal representations of their physical environments during spatial exploration. Rather than acquiring a rigid string of turn-left and turn-right motor habits, the animal’s central nervous system operates like an internal map room, continuously organizing environmental relationships into a comprehensive field map. In Tolman’s view, the brain synthesizes incoming perceptual cues into an overarching, relational, and metric internal structure that preserves the geometric relationships between landmarks, paths, and goals.

The definitive empirical demonstration of this principle arrived via Tolman’s classic latent learning experiments. When rodents were allowed to explore an unrewarded, intricate multi-path maze over multiple days without food reinforcement, orthodox behaviorism predicted that no learning could possibly occur, as no reinforcement was present to stamp in S-R associations. Yet, the precise moment a food reward was introduced to a specific goal box, these previously unrewarded rats exhibited an instantaneous, precipitous drop in navigational errors, matching or exceeding the performance of cohorts that had received daily reinforcement. The animals had clearly acquired extensive knowledge regarding the maze’s spatial topology in the complete absence of reward, deploying this latent cognitive map as soon as motivation dictated a goal-directed trajectory.

Tolman cemented this conceptual paradigm through his “sunburst” maze experiments. Rodents trained to navigate a winding, tortuous path toward a goal were abruptly presented with an open circular arena containing an array of novel, radiating directional paths. Instead of perseverating on the learned initial motor turns (the behaviorist prediction), the rodents overwhelmingly selected the novel, unlearned radial path that pointed directly along a straight vector toward the physical location of the hidden reward box. This calculation of novel shortcuts provided compelling evidence of a genuine metric relational schema: an internal representation capable of deriving vectors across unvisited territory, forever divorcing spatial navigation from simplistic associative chains.

1.2 Mid-Century Hippocampal Neuropsychology and Scoville-Milner Insights

While Tolman established the psychological reality of the cognitive map, the neuroanatomical locus of this internal metric remained completely unknown for decades. The search for the physical engram of spatial and relational knowledge gained unprecedented momentum in the mid-1950s through clinical neuropsychology. In 1957, neurosurgeon William Beecher Scoville and neuropsychologist Brenda Milner published their landmark case study of patient H.M. (Henry Molaison), who had undergone bilateral medial temporal lobe resection to alleviate intractable epilepsy. The surgical excision excised substantial portions of the amygdala, entorhinal cortex, and the anterior two-thirds of the hippocampus bilaterally.

The post-surgical consequence for H.M. was devastating and permanent: a profound, dense global anterograde amnesia. While H.M.’s intellectual capacity, sensory perception, language faculties, and short-term working memory remained intact, he lost the ability to convert any ongoing conscious experience into enduring long-term declarative memory. He could no longer form memories of autobiographical episodes, historical facts, or new acquaintances. Milner’s systematic testing demonstrated a critical dissociation: while declarative and episodic memory consolidation was thoroughly destroyed, H.M. retained the capacity to acquire new motor skills, such as mirror-tracing tasks, despite possessing no conscious recollection of ever having engaged in the training sessions.

These clinical findings led neuropsychologists to conclude that the hippocampus functioned primarily as a general-purpose declarative memory engine, responsible for consolidating multimodal episodic and semantic traces across the neocortex. However, this human neuropsychological model collided with a frustrating empirical paradox in animal laboratories. When researchers made extensive bilateral surgical or electrolytic lesions of the hippocampus in rodents and primates, the animals failed to exhibit the devastating global amnesia observed in human patients. Lesioned rats demonstrated normal acquisition and retention on a vast battery of standard laboratory tasks, including visual discrimination, classical conditioning, olfactory habituation, and basic associative operant paradigms.

The failure of animal hippocampal lesions to mirror human declarative amnesia ignited fierce debate. Some theorists argued that the rodent hippocampus was functionally distinct from its primate counterpart, or that human amnesia reflected neocortical or parahippocampal damage rather than hippocampal loss per se. Others postulated that the rodent tasks employed were insufficiently complex to tap into declarative awareness. What was urgently needed was an overarching, evolutionarily grounded neurophysiological framework that could reconcile the exquisite preservation of basic associative conditioning with the catastrophic navigational and episodic memory failures documented across species.

1.3 The Convergence of Electrophysiology and Ethological Psychology

The breakthrough that resolved this cross-species impasse arose not from human lesion neurology, but from technical and conceptual innovations within rodent electrophysiology during the late 1960s. Pioneering neurophysiologists, notably James Ranck Jr. and Felix Strumwasser, developed techniques for fabricating and chronically implanting microscopic insulated wire electrodes into the deep cerebral structures of freely moving animals. Prior to these developments, neurophysiological recordings were restricted to anesthetized or rigidly head-restrained preparations. These immobility paradigms fundamentally limited investigators to analyzing passive sensory receptive fields or stereotyped motor responses to artificial, discrete stimuli.

Recognizing that neural representations of natural behavior could only be deciphered when an animal was unconstrained, John O’Keefe, working at University College London, perfected microdrive assemblies capable of lowering ultra-fine, varnished wires into the dorsal hippocampus of freely moving rats. Crucially, O’Keefe merged this delicate technical mastery with an ethological perspective. He rejected the artificial strictures of traditional operant conditioning chambers—where an animal sat motionless waiting for a tone to press a lever—and instead observed rodents engaging in innate, unconstrained ethological behaviors: sniffing, rearing, grooming, foraging, and exploring open fields and complex elevated mazes.

In this naturalistic context, the functional architecture of the dorsal hippocampus began to reveal itself. Rather than finding units tuned to generic sensory modalities (such as specific visual angles, pure auditory frequencies, or cutaneous pressures), O’Keefe observed something unprecedented: individual pyramidal neurons whose firing rate remained baseline-silent across vast swathes of an arena, only to burst into intense, sustained high-frequency firing whenever the animal navigated into a specific, circumscribed physical coordinate in the laboratory room. These serendipitous recordings showed that hippocampal unit activity was modulated by an abstract, relational spatial variable—the organism’s physical location in environmental space.

2. The 1978 Landmark Monograph: The Hippocampus as a Cognitive Map

2.1 Core Tenets of the O’Keefe and Nadel Treatise

Seven years after the initial electrophysiological discovery, John O’Keefe joined forces with cognitive psychologist Lynn Nadel to publish their definitive 1978 theoretical treatise, The Hippocampus as a Cognitive Map. Published by Oxford University Press, this monumental work remains one of the most cited and intellectually cohesive syntheses in modern neuroscience. Spanning comparative neuroanatomy, evolutionary theory, behavioral psychology, electrophysiology, and epistemology, the book presented a rigorous, comprehensive model asserting that the dorsal hippocampus provides the physical instantiation of Edward Tolman’s cognitive map.

O’Keefe and Nadel structured their thesis around a fundamental dichotomy in spatial orientation mechanisms, dividing animal navigation into two radically distinct operational modes: egocentric (body-centered) orientation and allocentric (world-centered) mapping. Egocentric orientation relies on what they designated as Taxon systems (guidance toward single landmarks) and Route systems (fixed sequences of motor turns based on sensory stimuli encountered in real time). These egocentric frameworks are mediated by extra-hippocampal networks, including the striatum, cerebellum, and primary sensory cortices. In stark contrast, the hippocampus constitutes the Locale system: a dynamic, unified, allocentric metric framework that maps spatial relationships independently of the animal’s instantaneous physical position, heading, or motor behavior.

The authors argued that the locale system constructs an internal geometric coordinate grid within which objects, rewards, and historical events are embedded. This allocentric matrix grants the organism an internal representation of absolute space. Consequently, if familiar paths are blocked, or if an animal is dropped into a completely novel quadrant of a known environment, the hippocampal locale system can instantly calculate novel trajectories, identify shortcuts, and infer spatial relationships without requiring previous motor experience along that specific path. O’Keefe and Nadel extended this concept evolutionarily, proposing that this spatial coordinate system serves as the universal neural scaffolding upon which generalized memory, contextual binding, and ultimately human declarative and linguistic capabilities evolved.

2.2 Philosophical Underpinnings: Kantian Synthetic A Priori Intuitions

What uniquely elevated The Hippocampus as a Cognitive Map above ordinary neurobiological literature was its deliberate, sophisticated integration of Western epistemology. In the extensive philosophical prologue of the monograph, O’Keefe and Nadel traced the conceptual evolution of spatial theory, contrasting the empiricist models of John Locke, George Berkeley, and David Hume with the critical rationalism of Immanuel Kant.

The empiricist tradition asserted that the concept of space is an abstract, generalized construct synthesized purely a posteriori from discrete sensory experiences. In this view, spatial understanding is built incrementally by associating repeated sensations of vision, touch, and kinesthesia. O’Keefe and Nadel rejected the empiricist perspective as neurobiologically untenable. They argued that an organism cannot extract metric spatial relationships from raw, unstructured sensory inputs unless it already possesses an antecedent computational framework capable of organizing those inputs into spatial dimensions.

Instead, O’Keefe and Nadel explicitly grounded their neural theory in Immanuel Kant’s doctrine of space as an a priori intuition, as articulated in the Critique of Pure Reason (1781). Kant maintained that space is not an empirical concept derived from external experience, but an inherent, subjective form of sensibility: an internal lens or operating system that precedes all sensory perception and makes the comprehension of objects possible. O’Keefe and Nadel biologicalized Kant’s synthetic a priori. They proposed that the dorsal hippocampus is the physical organ of this Kantian intuition. Through millions of years of evolutionary adaptation, the mammalian genome encodes a pre-wired, self-organizing neural architecture that provides an innate coordinate framework. Sensory experiences do not create the spatial map; rather, sensory impressions are projected onto, and contextualized by, an antecedent neural spatial geometry embedded within the hippocampal microcircuitry.

2.3 Immediate Reception, Controversy, and Paradigm Shift

The publication of the 1978 monograph provoked an immediate and contentious scientific debate. The behavioral neuroscience community was entrenched in classical associative learning theory, which viewed all complex animal behaviors through the lens of stimulus-response-reinforcer matrices. Prominent behavioral psychologists attacked the cognitive map theory as unscientific, viewing the attribution of an internal, allocentric geometric map to rodents as an unnecessary regression into mentalism and anthropomorphic teleology.

Simultaneously, cognitive neuropsychologists specializing in human memory voiced intense skepticism. Researchers such as David Olton proposed alternative models, arguing that the hippocampus functioned not as a dedicated spatial mapping engine, but as a temporary workspace for “working memory” regardless of the spatial or non-spatial nature of the stimuli. Shortly thereafter, Howard Eichenbaum and Neal Cohen championed the “relational memory” hypothesis, contending that the apparent spatial specialization of rodent place cells was an experimental artifact of spatial navigation tasks. They posited that the hippocampus is fundamentally an arbitrary relational processing network that binds together any co-occurring stimuli across space and time, with physical space representing merely one dimension of a generalized declarative memory space.

Despite this fierce pushback, the predictive utility of O’Keefe and Nadel’s model unleashed a massive methodological revolution. The 1978 text provided exact, falsifiable predictions regarding lesion deficits, behavioral choices, and electrophysiological properties. It directly inspired the development of revolutionary behavioral assays designed specifically to isolate allocentric navigation from egocentric habits, most notably the Morris water maze and Olton’s radial arm maze. As empirical data rapidly accumulated showing that hippocampal damage selectively obliterated allocentric topological orientation while leaving complex associative learning pristine, the cognitive map theory transitioned from a radical hypothesis to one of the foundational pillars of modern systems neuroscience.

3. Electrophysiological Foundations: The Discovery of Place Cells

3.1 O’Keefe and Dostrovsky (1971) and Initial Discoveries

The empirical foundation of the cognitive map theory was first laid in a brief, revolutionary 1971 paper by John O’Keefe and his student Jonathan Dostrovsky, entitled “The Hippocampus as a Spatial Map. Preliminary Evidence from Unit Activity in the Freely Moving Rat”. Operating with rudimentary custom-built recording devices and audio monitors that converted neural action potentials into audible static clicks, they monitored the activity of individual single units in the CA1 region of the dorsal hippocampus as rats roamed freely within an open rectangular platform.

O’Keefe and Dostrovsky observed that neurons in the hippocampal formation segregated into distinct physiological profiles. One class of units, which they designated as “theta cells” (subsequently identified as fast-spiking, GABAergic inhibitory interneurons), fired at sustained high frequencies (typically 20–50 Hz) phase-locked to continuous macroscopic sinusoidal field potentials. The second, more numerous class of cells—now recognized as excitatory glutamatergic pyramidal neurons—exhibited low spontaneous baseline firing rates, often remaining silent for minutes at a time. However, these complex-spike pyramidal units suddenly fired prolonged bursts of action potentials whenever the rat traversed a particular spatial boundary of the testing arena.

To confirm that these firing patterns were not sensory artifacts, O’Keefe systematically manipulated the sensory environment. He tested whether the firing of these units was driven by specific tactile inputs from the paws, olfactory traces left on the maze surface, localized auditory reverberations, or specific angles of ambient illumination. In control trials, he rotated the maze apparatus, swiped clean the walking surfaces to eliminate scent trails, turned off all laboratory lights, and introduced novel objects into the space. The units maintained their rigid spatial selectivity regardless of these changes, continuing to discharge whenever the rat occupied the specific physical coordinate defined relative to the broader room boundaries. O’Keefe designated these remarkable neurons place cells, and the circumscribed physical area evoking their firing as the cell’s place field.

3.2 Characteristics of Place Fields

Subsequent decades of quantitative microelectrode recordings have comprehensively detailed the functional parameters of place fields. A classical place field is an allocentrically defined perimeter within an environment where the firing rate of an individual pyramidal neuron rises from near-zero baseline levels (<0.1 Hz) to robust, concentrated discharge frequencies ranging from 5 Hz to upwards of 30 Hz. The place field is fundamentally stable; once a place cell forms a receptive field in a familiar environment, that exact same neuron will reliably fire in the identical spatial coordinate across days, weeks, and even months, provided the structural environment remains unchanged.

However, when an animal is transitioned from one distinct environment to another (for example, from a square grey arena to a circular black cylinder), the hippocampal ensemble undergoes a dynamic, non-linear transformation known as remapping. Electrophysiologists recognize two primary variants of this phenomenon:

• Rate Remapping: Occurs when the global geometry and coordinate framework of an environment remain constant, but local features (such as color, wall texture, or sensory cues) undergo moderate alterations. Under rate remapping, the place field coordinates remain physically pinned to the same spatial locations, but the peak firing rates within those fields undergo dramatic, independent modulations.
• Global Remapping: Occurs when the animal is introduced to an entirely novel context, or when the fundamental geometric boundaries of an arena are drastically shifted. In global remapping, the active place cell ensemble undergoes a complete, stochastic orthogonalization. Cells that fired in the original environment fall silent; previously dormant cells activate in novel locations; and active cells relocate their place fields to completely unrelated coordinates across the new arena.

The morphology and boundaries of place fields are profoundly governed by the macroscopic physical geometry of the enclosure. In a classic series of experiments, O’Keefe and Burgess manipulated the dimensions of rectangular recording boxes by independently extending or contracting the walls. They discovered that place fields stretched, duplicated, or compressed systematically in response to the repositioning of specific boundary walls, demonstrating that place cell firing is mathematically derived from an environmental boundary-referencing framework.

3.3 Theta Rhythm and Phase Precession

The firing of hippocampal place cells is temporally orchestrated by an intense, continuous macroscopic oscillation known as the theta rhythm. Ranging between 6 to 10 Hz in active, locomoting rodents, this rhythmic local field potential (LFP) is generated by pacemaker circuits within the medial septum and vertical limb of the diagonal band of Broca, which project via cholinergic and GABAergic fibers to hippocampal interneuronal networks. The theta oscillation provides a dynamic clocking mechanism that synchronizes hippocampal pyramidal assemblies.

In 1993, John O’Keefe and his colleague Michael Recce discovered a remarkable phenomenon operating at the intersection of spatial encoding and neural timing: theta phase precession. Prior to this discovery, neuroscientists viewed place cells through a pure rate-coding lens: the closer an animal was to the center of a place field, the faster the cell fired. O’Keefe and Recce demonstrated that place cells also employ an exquisite, high-resolution temporal code. As a rat enters the peripheral boundary of a place field, the unit fires its first action potentials at a specific, late phase of the ongoing 8 Hz theta wave cycle (near the peak of the oscillation).

As the rat continues along its trajectory and traverses through the physical center of the field, the timing of the action potentials systematically advances to earlier and earlier phases of the theta cycle. By the time the animal reaches the exit boundary of the place field, the spikes discharge at the earliest phase of the theta wave (near the trough), completing an approximate 360-degree phase precession across the traverse of the field. Phase precession fundamentally altered theoretical neuroscience. It proved that the precise timing of action potentials relative to an internal network clock carries an order of magnitude more spatial information than firing rate alone, allowing downstream decoders to reconstruct an animal’s exact millimeter-level coordinate within a single 120-millisecond theta cycle.

Phase precession provides a crucial solution to a central problem in neurobiology: the temporal compression paradox. The biochemical induction of Spike-Timing-Dependent Plasticity (STDP) requires that the presynaptic and postsynaptic neurons fire action potentials within a tight temporal window of 10 to 30 milliseconds. However, when an animal physically walks through an environment, it takes several seconds to travel between adjacent spatial locations. By virtue of phase precession, this behavioral timescale (spanning seconds) is systematically compressed into a single theta cycle (120 milliseconds). Place cells representing consecutively traversed locations fire sequentially within every theta wave, perfectly separated by 10-to-20-millisecond intervals. This temporal compression allows classical STDP mechanisms to permanently bind sequentially activated spatial locations into lasting memory traces.

4. Anatomical Architecture and Microcircuitry of the Hippocampus

4.1 The Trisynaptic Circuit and Information Flow

To understand how the hippocampus computes the allocentric cognitive map, one must trace the precise cytoarchitectonic wiring of the medial temporal lobe. Spatial and sensory information is synthesized across broad neocortical association areas and funneled directly into the parahippocampal gyrus, specifically the lateral and medial entorhinal cortices (LEC and MEC). From the entorhinal cortex, information enters the canonical, unidirectional feedforward loop classically designated as the trisynaptic circuit, first conceptualized by the great neuroanatomist Santiago Ramón y Cajal and electrophysiologically mapped by Per Andersen.

The primary conduit of cortical input into the hippocampus is the perforant path. Originating primarily from stellate and pyramidal neurons located within Layer II of the entorhinal cortex, these axons cross the subicular cleft to make dense, excitatory glutamatergic synapses upon the dendritic spines of the granule cells in the dentate gyrus (the first synaptic station). The dentate gyrus processes these inputs and sends exceptionally thick, unmyelinated axons known as mossy fibers to synapse upon the massive complex spines (thorny excrescences) of the pyramidal neurons in the CA3 (Cornu Ammonis 3) subfield, forming the second synaptic station.

The CA3 pyramidal neurons, in turn, cast off branched axonal pathways. While some collateralize locally, their primary forward projections—the Schaffer collaterals—ascend through the stratum radiatum to synapse upon the apical dendrites of pyramidal neurons within the CA1 subfield, completing the classical trisynaptic loop. CA1 then projects robustly to the subiculum, which provides the major structural outflow of the hippocampal formation. The subiculum and CA1 project backward to the deep layers (Layers V and VI) of the entorhinal cortex, completing an interconnected loop that redistributes processed spatial and contextual representations back out to the neocortex.

4.2 CA3 as an Autoassociative Network

The computational engine of the hippocampal trisynaptic circuit resides within the unique microcircuitry of the CA3 subfield. In addition to projecting forward to CA1 via the Schaffer collaterals, CA3 pyramidal neurons possess massive, highly branched recurrent collaterals. Each individual CA3 pyramidal cell emits an extensive axonal arbor that projects backward to make direct excitatory synaptic contacts with thousands of neighboring CA3 pyramidal neurons throughout the longitudinal axis. In his visionary 1971 mathematical treatise, David Marr hypothesized that this anatomical architecture functions as an autoassociative network.

Autoassociative networks possess the unique computational capacity to perform pattern completion. When an organism explores an environment, the recurrent excitatory synapses between co-active CA3 pyramidal neurons undergo massive, coordinated long-term potentiation, binding these cells into a robust, mutually stabilizing Hebbian cell assembly. Subsequently, if the animal is returned to that environment under degraded, ambiguous, or sensory-deprived conditions (such as complete darkness or with key landmarks removed), an incomplete fragment of the original sensory input arrives via the perforant path. This partial sensory cue activates a fraction of the original CA3 ensemble; the powerful recurrent collaterals instantly transmit excitatory drive across the entire recurrent web, igniting the remaining silent neurons and fully retrieving the complete, pristine global cognitive map of the space.

In modern computational terms, the recurrent architecture of CA3 is modeled as a continuous attractor neural network (CANN). The energetic landscape of this network forms a smooth, low-dimensional manifold where stable patterns of localized firing (activity “bumps”) correspond to specific locations in the physical environment. Even when external sensory drive fluctuates or momentarily vanishes, the intrinsic recurrent feedback within CA3 sustains the continuous location representation on the manifold, protecting the internal cognitive map from drifting into chaos.

4.3 Dentate Gyrus and CA1 Functional Specialization

While CA3 is optimized for pattern completion, its powerful recurrent collateral network creates an inherent computational danger: if multiple similar environments project into CA3, the recurrent excitatory web risks collapsing these distinct inputs into a single, catastrophic blend, leading to catastrophic interference. The hippocampal microcircuitry prevents this collapse through the specialized computational operations of the dentate gyrus (DG).

The dentate gyrus acts as the principal gatekeeper of the hippocampus, executing pattern separation. The dentate gyrus contains an order of magnitude more granule cells than there are projecting stellate cells in Layer II of the entorhinal cortex. Crucially, the baseline activity of dentate granule cells is governed by powerful, hyperpolarizing GABAergic feedforward and feedback inhibition from local interneurons (such as basket cells and chandelier cells), rendering the dentate gyrus exceptionally sparse in its output: at any given moment, less than 2% of granule cells are firing action potentials. When two highly similar, overlapping sensory inputs arrive from the entorhinal cortex, the dentate gyrus orthogonalizes them, projecting entirely distinct, non-overlapping subsets of mossy fibers onto CA3. Furthermore, the dentate gyrus is one of only two structures in the adult mammalian brain that undergoes continuous adult neurogenesis. Adult-born immature granule cells exhibit hyper-excitable physiological states that dynamically integrate temporal tags into spatial codes, ensuring that experiences occurring at different times in identical spaces remain segregated.

At the downstream pole of the circuit, the CA1 subfield acts as a sophisticated comparator. Unlike CA3, CA1 lacks recurrent collateral connections, functioning instead as a feedforward feed-through network. CA1 uniquely receives two radically distinct inputs: the processed, pattern-completed, or pattern-separated representation channeled indirectly from CA3 via the Schaffer collaterals, and direct, unprocessed neocortical sensory inputs originating from Layer III of the entorhinal cortex via the temporoammonic pathway. This dual-input convergence allows CA1 to act as a match/mismatch detector, comparing the internalized prediction generated by the CA3 cognitive map against the real-time sensory reality streaming directly from the neocortex.

Finally, these physiological and computational operations are strictly mapped along the longitudinal (dorsoventral) axis of the hippocampus:

• Dorsal Hippocampus (Septal Pole): Devoted to high-resolution, metric spatial mapping. Dorsal place cells have exceptionally compact, sharp, and highly granular place fields (spanning 20–50 cm in diameter), providing precise spatial coordinates.
• Ventral Hippocampus (Temporal Pole): Place fields expand dramatically in scale, often encompassing entire rooms or continuous tracts spanning several meters. The ventral pole is intimately interconnected with the basolateral amygdala, prefrontal cortex, and nucleus accumbens, transforming metric spatial maps into broad contextual representations laced with emotional valence, fear, anxiety, and motivational significance.

5. Spatial Reference Frames: Egocentric versus Allocentric Coding

5.1 Taxon and Route Systems versus Locale Systems

A central theoretical pillar of O’Keefe and Nadel’s 1978 framework was the rigorous dissociation between egocentric and allocentric spatial navigation. They formalized this distinction by contrasting Taxon and Route navigation systems against the hippocampal Locale system.

The Taxon system is subdivided into two primary navigational behavioral strategies:

• Guidance: The simplest form of orientation, wherein an animal directs its movement toward or away from a discrete, salient sensory cue (e.g., swimming directly toward an elevated, visible platform, or running toward a flashing light). This strategy requires only egocentric sensory alignment; the animal merely reduces the visual or auditory distance between its body and the beacon.
• Orientation / Motor Chaining: A form of navigation wherein the animal executes a fixed string of proprioceptive and kinesthetic motor turns (e.g., “turn left at the first junction, run forward 10 strides, turn right”). These route strategies are inherently egocentric, rigidly tied to the organism’s point-of-view and body axis. They are mediated predominantly by the dorsal striatum (caudate/putamen), premotor cortices, and cerebellum.

The fatal vulnerability of Taxon and Route systems is their extreme fragility. If an environmental obstacle blocks an established route, or if an animal is displaced unexpectedly into a novel sector of an arena, egocentric chains break down catastrophically, forcing the organism into random trial-and-error behaviors. In direct contrast, the hippocampal Locale system constructs an absolute allocentric metric framework. Spatial coordinates are established not relative to the animal’s physical body, but relative to the static, external geometric configuration of the global environment. Because the locale map internalizes the distance and angle relationships between all distal cues and boundaries, an animal operating within this allocentric system can be dropped into unvisited coordinates and instantly orient itself, calculate detours around novel roadblocks, and synthesize novel trajectories toward unvisited goals.

Decades of behavioral dissociation studies confirm this anatomical partition. Animals suffering selective, complete neurotoxic lesions of the dorsal hippocampus can easily acquire visual cue-guidance tasks (Taxon) and learn highly stereotyped sequence turns in a cross maze (Route). However, the moment these animals are required to execute allocentric navigation—such as finding a hidden platform submerged beneath opaque water from varying release points—they exhibit persistent spatial navigation deficits.

5.2 Coordinate Transformations in Parieto-Hippocampal Networks

The physical reality of an organism’s sensory apparatus is fundamentally egocentric. Photons strike the retina in a two-dimensional, retinocentric frame of reference; sounds strike the tympanic membranes in a head-centered frame; and mechanoreceptive inputs enter the somatosensory cortex in a body-centered coordinate frame. A central question in neurophysiology is how these fleeting, viewer-dependent egocentric sensory signals are transformed into the static, view-invariant allocentric coordinate map observed in the dorsal hippocampus.

This conversion is computed via a hierarchically organized sensory-parieto-retrosplenial-hippocampal network. Visual, tactile, and kinesthetic inputs are first integrated within the posterior parietal cortex (PPC). PPC neurons encode spatial information using body-centered egocentric vectors, mapping the locations of obstacles, boundaries, and objects relative to the animal’s eyes, head, and trunk. However, PPC projections terminate heavily in the retrosplenial cortex (RSC), an intermediate brain area that serves as the central coordinate-transformation hub.

The retrosplenial cortex contains a dual population of neurons: units that fire in response to egocentric positions, units that fire in response to allocentric boundaries, and an enigmatic class of cells that simultaneously code both reference frames. Working in close reciprocal dialogue with the postsubiculum and the anterior thalamic nuclei—which provide a continuous head-direction vector—the retrosplenial cortex applies vector rotation algorithms. It systematically converts egocentric sensory observations (e.g., “wall 20 centimeters to my physical left”) into allocentric geometric statements (e.g., “wall oriented along the western boundary of the room”). These view-invariant allocentric vectors are then projected into the entorhinal cortex and dorsal hippocampus, updating and anchoring the internal cognitive map in real time.

5.3 Flexible Navigation and Shortcut Computation

The defining computational property of a true metric cognitive map is its capacity to compute novel trajectories—specifically, the execution of shortcuts and dynamic detour behavior. In a rigid associative network or an egocentric motor-chaining model, an animal can only navigate along physical paths that it has previously traversed and reinforced. To navigate from Point A to Point C via an unvisited diagonal shortcut requires calculating a novel vector that has never been physically experienced.

O’Keefe and Nadel asserted that the hippocampal place cell ensemble provides precisely the metric coordinate scaffolding required for this vector calculation. Computational models, such as those pioneered by Neil Burgess and colleagues, illustrate how this computation operates. When an animal explores an environment, place cells fire sequentially, establishing continuous synaptic connectivity profiles across the hippocampal network. If the animal is situated at Point A and is motivated to reach a known reward location at Point C, the internal map activates the place cell assembly representing the current location (A) while simultaneously recalling the goal ensemble (C). By reading out the differential activation across this allocentric topological manifold, downstream executive circuits—particularly the medial prefrontal cortex—can compute the shortest Euclidean distance and calculate an unlearned, direct directional vector.

Empirical evidence for this allocentric shortcut capability is abundant. When rodents trained in complex mazes encounter unexpected barriers that block their habitual routes, intact animals do not blindly collide with the barrier or attempt to repeat the habit; they immediately turn and select the most direct alternative branching path that bypasses the blockage to intersect the goal vector. When the dorsal hippocampus is pharmacologically inactivated with GABA-A receptor agonists (such as muscimol), this flexible detour and shortcut capability evaporates completely. Muscimol-infused animals regress to rigid, repetitive egocentric movements, blindly running down the blocked corridor or wandering in aimless circles, validating O’Keefe and Nadel’s postulate that flexible topological problem solving requires an intact hippocampal locale system.

6. Beyond Pure Space: Non-Spatial Modulations in Hippocampal Coding

6.1 Olfactory, Auditory, and Visual Feature Coding

Although O’Keefe and Nadel initially formulated their theory around an absolute geometric coordinate framework, subsequent decades of recording revealed that the hippocampal cognitive map is not an empty, sterile spatial grid. Instead, it is an integrated representational manifold that binds rich sensory, perceptual, and episodic features directly onto the underlying spatial scaffold.

In rodents, who are nocturnal and fundamentally olfactory-driven creatures, the firing of hippocampal place cells is heavily modulated by chemical cues. In experiments where specific odorants are diffused into distinct quadrants of an arena, place fields rapidly become anchored to these olfactory gradients. Furthermore, in paradigms developed by Howard Eichenbaum and colleagues, animals were trained to sniff a series of distinct odorants presented at specific spatial ports. Pyramidal neurons in CA1 did not simply code where the animal was standing; they exhibited complex item-in-context coding, discharging maximally only when a specific, individual odor was presented within a specific physical location.

A striking demonstration of non-spatial sensory coding within the hippocampal mapping engine was provided by Dmitriy Aronov, Rhino Nevers, and David Tank in 2017. Rats were trained to manipulate a joystick to navigate through a continuous auditory frequency spectrum (a “sound sweep” ranging from 2 kHz to 40 kHz) to match an auditory target for a water reward. Remarkably, as the rats traversed this continuous non-spatial acoustic spectrum, dorsal CA1 and CA3 pyramidal neurons exhibited discrete, localized receptive fields tuned to specific sound frequencies. These neurons fired in narrow frequency bands precisely analogous to the way classic place cells fire in narrow physical coordinates. The hippocampus had co-opted its internal metric mapping architecture to construct a one-dimensional “cognitive map” of an auditory continuum, confirming that the spatial mapping circuitry of the hippocampus can represent any continuous, multi-dimensional relational variable.

6.2 Task Demands, Goals, and Reward Modulation

The firing fields of hippocampal pyramidal cells are dynamically reshaped by an animal’s cognitive state, behavioral intent, and the distribution of biological rewards. If the cognitive map were a purely passive, invariant geometric mirror of external space, place fields would distribute themselves randomly and uniformly across any available physical floor space. However, empirical recordings consistently reveal marked departures from spatial homogeneity.

When animals are trained on goal-directed navigation tasks, such as finding a hidden reward in an open arena or alternating choices on a maze, place fields undergo profound goal-directed clustering. An abnormally high percentage of active place fields skew and concentrate in immediate proximity to the reward delivery sites, physical escape platforms, or critical decision junctions. The spatial metric becomes distorted to over-represent areas of high biological valence, essentially expanding the neural resolution of the map where the computational stakes are highest.

Furthermore, in alternating T-maze or continuous W-maze paradigms, place cells exhibit what is termed prospective and retrospective coding (or “trajectory-dependent split-cell coding”). When a rat traverses the central stem of a T-maze—a path physically identical on every single trial—an individual place cell will fire robustly if the animal is planning to turn left at the choice point, but remain completely silent if the animal is planning to turn right. Even though the rat occupies the exact same physical coordinates, with identical sensory inputs and identical motor kinematics, the cell’s firing is split based on the animal’s future intended trajectory (prospective coding) or the location from which it just departed (retrospective coding). The hippocampal map is inherently trajectory-dependent, integrating instantaneous position with past history and future intentions.

6.3 Concept Cells and Semantic Abstraction in the Human Brain

The ultimate validation of the cognitive map theory as an evolutionary precursor to generalized conceptual thought arrived through single-unit recordings in the human brain. Neurosurgeon Itzhak Fried and neuroscientist Rodrigo Quian Quiroga recorded from individual neurons in the medial temporal lobe (hippocampus, entorhinal cortex, and amygdala) of neurosurgical patients undergoing invasive intracranial monitoring for intractable epilepsy.

These studies revealed the existence of sparse, highly abstract units popularly designated as concept cells (or the “Jennifer Aniston neuron”). These human medial temporal units fire selectively and invariant to completely disparate sensory representations of a single, specific concept. An individual hippocampal cell would fire bursts of action potentials when the patient was shown a color photograph of a specific celebrity, a line drawing of that person, the individual’s name typed out as text, or even the sound of the name spoken aloud. These neurons strip away all low-level perceptual features to encode the abstract semantic and relational essence of a unique entity.

In recent years, cognitive neuroscientists leveraging high-resolution functional Magnetic Resonance Imaging (fMRI) have shown that the human hippocampal-entorhinal network uses this metric architecture to map purely abstract, non-spatial multidimensional spaces. In landmark studies by Raymond Dolan, Timothy Behrens, and colleagues, human participants navigated through virtual spaces where items varied along abstract dimensions (such as the neck length and leg length of cartoon birds, or social dominance and trustworthiness hierarchies within a simulated social network). Human fMRI scans revealed the exact same metric codes (such as grid-like and place-like activation signatures) navigating through these abstract semantic and social spaces as are observed during physical spatial navigation. The cognitive map theory had come full circle: the hippocampal metric system evolved not merely to track physical terrain, but to navigate the multidimensional geometric landscapes of human thought, memory, and semantic concepts.

7. Synaptic Plasticity, Long-Term Potentiation, and Map Formation

7.1 Long-Term Potentiation (LTP) in the Perforant Path and Schaffer Collaterals

A cognitive map that cannot be updated, stabilized, and permanently stored would be biologically useless. The physiological mechanism that links the real-time firing of place cells to enduring spatial memory is synaptic plasticity, epitomized by the phenomenon of Long-Term Potentiation (LTP). First discovered by Terje Lømo in the rabbit dentate gyrus and published with Timothy Bliss in 1973 (“Long-lasting potentiation of synaptic transmission in the perforant path of the anaesthetized rabbit to stimulation of the perforant path”), LTP demonstrated that high-frequency electrical stimulation of an excitatory pathway produces a persistent, long-lasting increase in synaptic efficacy.

LTP throughout the trisynaptic circuit—particularly at the Schaffer collateral-CA1 pyramidal cell synapse—is heavily dependent on the activation of the N-methyl-D-aspartate (NMDA) receptor. Under basal, resting conditions, the ionotropic pore of the NMDA receptor is physically blocked by an extracellular magnesium ion ($Mg^{2+}$). Depolarization of the postsynaptic membrane via alpha-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid (AMPA) receptors electrostatically expels the $Mg^{2+}$ ion, allowing an influx of extracellular calcium ($Ca^{2+}$) into the dendritic spine.

The causal link between NMDA-dependent LTP and the stability of the hippocampal cognitive map was definitively demonstrated in landmark studies utilizing targeted pharmacology and transgenic mouse models. Richard Morris and colleagues showed that infusing the competitive NMDA receptor antagonist AP5 (2-amino-5-phosphonopentanoic acid) into the cerebral ventricles of rats did not abolish existing place fields; place cells still fired when an animal was placed in a familiar environment. However, AP5 completely prevented the stabilization and long-term retention of newly formed place fields in novel environments. In 1996, Susumu Tonegawa and colleagues generated a subfield-restricted knockout of the essential NMDA receptor subunit NR1 specifically within CA1 pyramidal cells (CA1-NR1 KO mice). These animals completely lacked Schaffer collateral LTP. When recorded in novel arenas, their place cells formed fragile, uncoordinated place fields that remap chaotically from one session to the next. The cognitive map had lost its synaptic glue.

7.2 Molecular Cascades of Map Consolidation

The transformation of transient, dynamic place cell discharges into an enduring, structurally stable cognitive map requires a cascade of intracellular signaling pathways that bridge the postsynaptic density to the cell nucleus:

1. Calcium Influx and Enzymatic Activation: Calcium entering through the open pore of NMDA receptors binds to calmodulin, forming a complex that activates Calcium/Calmodulin-Dependent Protein Kinase II (CaMKII). Autophosphorylated CaMKII translocates directly to the postsynaptic density, phosphorylating AMPA receptors to increase their single-channel conductance and driving the insertion of novel GluA1-containing AMPA receptors into the synaptic membrane.
2. Transcriptional Cascades: Sustained synaptic activation triggers the MAPK/ERK (Mitogen-Activated Protein Kinase) signaling cascade, which translocates into the nucleus to phosphorylate the transcription factor CREB (cAMP-Response Element-Binding Protein). Phosphorylated CREB initiates the transcription of immediate early genes (IEGs) critical for permanent structural remodeling, including Arc (Activity-Regulated Cytoskeleton-Associated Protein) and c-Fos.
3. Structural Spine Remodeling: Arc translocates directly back to the active dendritic spines, regulating actin cytoskeletal dynamics to structurally enlarge the spine head. New synaptic boutons and perforated postsynaptic densities are synthesized, converting a temporary, electrophysiologically induced place field into a hardwired structural connection within the CA3-CA1 network.

Visualizing these molecular cascades in real time has revealed that the activation of immediate early genes mirrors the animal’s cognitive mapping behavior. When an animal explores a novel environment, Arc and c-Fos exhibit rapid, robust upregulation specifically within the discrete ensemble of hippocampal place cells that are actively engaged in mapping that specific spatial coordinate. If these molecular pathways are blocked via protein synthesis inhibitors (such as anisomycin), the initial formation of place fields remains intact, but their retention across 24-hour intervals is destroyed: the cognitive map fades, and the animal treats the familiar environment as entirely novel.

7.3 Experience-Dependent Map Refinement and Remapping

The construction of a cognitive map is not an instantaneous, static event; it is an experience-dependent process that unfolds across distinct temporal phases. When an animal steps into a completely novel, unmapped environment, a functional hippocampal representation forms with remarkable speed. Within 2 to 5 minutes of novel exploration, dorsal CA1 and CA3 place cells begin to fire in localized fields. The baseline geometry of the space is drafted almost immediately.

However, over the course of hours, days, and repeated exposures, this rough draft undergoes significant experience-dependent refinement. Longitudinal electrophysiological tracking has shown that the spatial tuning of place fields sharpens progressively: the perimeters of the fields contract, the peak firing rate within the center increases, the baseline out-of-field noise drops to zero, and the directional selectivity of the fields deepens. The network fine-tunes its synaptic weights, carving a high-fidelity continuous attractor manifold out of the initially loose synaptic connections.

This structural stabilization also governs the dynamics of hysteresis—the network’s capacity to preserve an existing internal map state despite ambiguous, fluctuating sensory inputs. If an environment is slowly, incrementally deformed (for example, gradually altering the aspect ratio of a rectangular arena into a square over multiple intermediate steps), the hippocampal place cell ensemble will stubbornly preserve its original firing configuration, showing memory persistence. Only when a critical sensory threshold is crossed does the network suddenly snap into an entirely new, orthogonalized map state. This non-linear transition illustrates that the cognitive map is not a passive sensory slave, but a self-sustaining internal model that resists perturbation until environmental changes dictate a fundamental contextual shift.

8. The Extended Spatial Navigation Circuit: Entorhinal and Subicular Networks

8.1 Head Direction Cells: The Neural Compass

A cognitive map cannot effectively guide navigation without an internal metric defining orientation. Just as a physical cartographic map requires a compass rose indicating North, South, East, and West, the hippocampal place cell system must be continuously informed of the organism’s directional heading in horizontal space. This orientation vector is computed by head direction (HD) cells, first discovered in the rat postsubiculum by James Ranck Jr., and comprehensively characterized by Jeffrey Taube, Robert Muller, and Ranck in 1990 (“Head-direction cells recorded from the postsubiculum in freely moving rats. I. Description and quantitative analysis”).

Head direction cells function as an internal, allocentric biological compass. An individual HD cell fires at a high sustained rate whenever the animal’s head is pointed in a specific azimuthal direction relative to the global environment, independent of the animal’s physical location in the room, its posture, or whether it is moving or standing still. If the animal points its head 45 degrees away from this preferred firing direction, the unit’s discharge rate drops to absolute silence. These cells are organized across an anatomical circuit encompassing the dorsal tegmental nucleus of Gudden, the lateral mammillary nucleus (LMN), the anterodorsal thalamic nucleus (ADN), the postsubiculum (PoS), and the retrosplenial cortex.

The generation of the head direction signal relies on the integration of self-motion angular velocity signals streaming from the vestibular semicircular canals. The lateral mammillary-anterodorsal thalamic network acts as a ring attractor that integrates this angular head velocity over time, continuously rotating a stable bump of neural activity across a circular manifold. This heading vector projects directly into the medial entorhinal cortex and the dorsal hippocampus. The functional coupling between the head direction system and the place cell system is absolute: if distal visual landmarks are rotated by 90 degrees, the preferred firing directions of HD cells rotate by exactly 90 degrees, and the place fields of hippocampal place cells rotate in perfect, lockstep synchrony. If the head direction system is lesioned, place fields lose their directional stability and degenerate into noisy, drifting patches.

8.2 Grid Cells and the Metric Scaffolding (Moser and Moser)

While O’Keefe and Nadel’s theory posited an internal Euclidean metric for the cognitive map, the exact nature of this metric framework remained a computational mystery until 2005. Place cells, with their single localized firing fields, were poorly equipped to measure absolute distance across broad, unvisited space; an animal would need an existing field to know a distance. The underlying metric was revealed when Edvard Moser, May-Britt Moser, and their students Marianne Fyhn, Sturla Molden, and Torkel Hafting recorded from the dorsomedial entorhinal cortex (MEC).

In a historic 2005 paper published in Nature (“Microstructure of a spatial map in the entorhinal cortex”), the Moser laboratory reported the discovery of grid cells. Unlike a hippocampal place cell, which fires in only one circumscribed location in an environment, a grid cell fires at multiple, regularly spaced locations across the entire surface of an arena. The spatial distribution of these multiple firing fields is stunningly geometric: the fields form a periodic, isotropic, triangular tessellation that tiles the entire 2D physical space in a perfectly regular hexagonal array.

Grid cells provide the universal, internal metric coordinate system of the mammalian brain. Every grid cell possesses three defining spatial parameters:

• Grid Scale (Wavelength): The physical distance separating the peaks of the hexagonal firing vertices. Grid scale is anatomically organized along a topographically rigid gradient running down the dorsoventral axis of the MEC. Grid cells located at the dorsal pole exhibit exceptionally tight, fine-grained scales (separated by roughly 30 cm), while cells recorded at progressively ventral coordinates exhibit expanding scales, reaching several meters in wavelength.
• Grid Orientation: The angular orientation of the hexagonal lattice relative to an external reference frame.
• Spatial Phase: The two-dimensional offset or coordinate alignment of the lattice peaks relative to an external origin point.

The discovery of grid cells provided the missing computational scaffolding for the O’Keefe-Nadel model. Theoretical and empirical models demonstrated that the single, localized firing field of a hippocampal place cell can be mathematically derived via linear summation of multiple overlapping grid cell inputs of varying scales and phases projecting from the MEC along the perforant path. In 2014, John O’Keefe, May-Britt Moser, and Edvard Moser were jointly awarded the Nobel Prize in Physiology or Medicine for their complementary discoveries of place cells and grid cells—the internal GPS and cognitive mapping engine of the brain.

8.3 Boundary Vector and Border Cells

A continuous metric map derived solely from path integration will inevitably accumulate drift and error over time unless it is constantly anchored to the physical perimeters of the environment. To address this computational imperative, Neil Burgess, Colin Lever, and John O’Keefe mathematically modeled the existence of theoretical units they termed Boundary Vector Cells (BVCs). They predicted that these neurons would fire whenever an environmental boundary (such as a drop-off, wall, or partition) was encountered at a specific distance and allocentric direction relative to the animal.

The empirical confirmation of this theoretical prediction arrived in 2008, when Trygve Solstad, Edvard Moser, and May-Britt Moser discovered border cells in the medial entorhinal cortex, and Colin Lever and colleagues identified them within the subiculum. Border cells fire continuously along one or more physical boundaries of an arena. For example, an individual border cell will fire intensely along the entire northern wall of an enclosure, regardless of whether that wall is 1 meter long or stretched to 3 meters long.

Border and boundary vector cells serve as the critical functional interface between the external sensory environment and the internal cognitive map. Whenever an animal physically approaches or brushes against a wall, border cells fire, delivering a powerful corrective signal to the continuous attractor networks of grid and place cells. This boundary interaction immediately resets the path integrator, scrubbing out cumulative metric errors and locking the grid and place fields securely to the environmental architecture.

9. Path Integration, Self-Motion Cues, and Sensory Recalibration

9.1 Mechanisms of Idiothetic Navigation

How does the hippocampal cognitive map maintain its spatial representation when external landmarks become invisible, such as when an animal navigates an underground burrow in complete darkness? The cognitive map survives sensory deprivation through a computational process known as path integration, or idiothetic navigation (dead reckoning).

Path integration is the mathematical process of continuously updating an internal position vector relative to an origin point by integrating self-motion signals over time. The brain relies on three primary sources of idiothetic information:

• Vestibular Cues: Linear accelerations sensed by the otolith organs (utricle and saccule) and angular accelerations detected by the semicircular canals, providing real-time data on head velocity and directional shifts.
• Proprioceptive Feedback: Kinesthetic signals generated by stretch receptors in muscles, tendons, and joints, signaling the mechanical speed and stride count of the limbs during locomotion.
• Motor Efference Copy: Collateral copies of motor commands dispatched from frontal motor cortices to the brainstem and spinal cord, alerting sensory and mapping networks to the amplitude of an intended movement before muscle contraction occurs.

Within the medial entorhinal cortex, specialized speed cells fire at rates linearly proportional to the animal’s instantaneous running speed. By multiplying this running speed signal with the heading vector provided by head direction cells, the MEC-hippocampal network continuously integrates velocity over time ($\vec{x}(t) = \vec{x}(0) + \int \vec{v}(t) dt$), shifting the internal place and grid representations through space in total darkness. However, path integration has an inherent physical limitation: because it integrates velocity signals over time, any minor noise or inaccuracy in the sensory stream is integrated into the calculation, causing metric error to compound exponentially over prolonged absences of external sensory landmarks.

9.2 Visual Cue Realignment and Landmark Correction

To counteract the inexorable drift of pure path integration, the mammalian brain relies on allothetic navigation—the continuous integration of external sensory landmarks, primarily distal visual cues. Visual landmarks serve as the absolute ground truth that recalibrates and phase-resets the internal path integrator.

The dynamic interplay between idiothetic self-motion cues and allothetic visual landmarks has been thoroughly dissected using cue-conflict paradigms in virtual reality and physical arenas. In these experiments, an animal runs on a spherical treadmill while navigating a visual virtual environment. Experimenters systematically decouple the physical distance the animal runs (idiothetic gain) from the rate at which the visual scenery streams past (optic flow gain), or covertly rotate distal visual landmarks relative to the physical geometry of the enclosure.

These paradigms demonstrate that when the conflict between idiothetic cues and visual landmarks is small (e.g., a 10-degree rotation of distal cues), allothetic vision completely dominates: place fields and grid lattices phase-shift and rotate in lockstep with the visual cues, recalibrating the path integrator. However, if the conflict exceeds a critical threshold (e.g., visual cues are rotated by 180 degrees or optic flow is accelerated to unnatural velocities), the hippocampal network undergoes a profound rupture: the internal map breaks away from the visual landmarks, falling back on internal idiothetic path integration and local physical tactile boundaries, or splitting into two independent, competing network representations.

9.3 Continuous Attractor Neural Networks (CANNs)

To mathematically explain how idiothetic velocity signals shift internal spatial representations without losing field stability, computational neuroscientists formulated Continuous Attractor Neural Networks (CANNs). Pioneered by Samsonovich, McNaughton, Zhang, and Fiete, CANN models conceptualize grid and place cell sheets as continuous, low-dimensional attractor manifolds.

In a 2D continuous attractor network of grid cells, the neurons are arranged on a virtual computational grid with toroidal (donut-shaped) periodic boundary conditions. The synaptic architecture of the network is defined by a local excitation, long-range inhibition (“Mexican-hat”) connectivity matrix:

• Neurons that represent adjacent spatial phases possess strong, reciprocal excitatory synaptic connections.
• Neurons that represent distant spatial phases send powerful inhibitory projections via local GABAergic interneurons.

This recurrent balance ensures that the network spontaneously settles into a localized “bump” of high-frequency action potential activity. To translate this activity bump across the neural sheet during physical locomotion, the network incorporates asymmetric feedforward projections modulated by the animal’s instantaneous velocity. Directional velocity signals (provided by conjugately tuned head-direction-by-speed cells) preferentially excite the leading edge of the activity bump while inhibiting the trailing edge. As the animal runs, the activity bump glides smoothly across the toroidal neural sheet. Because the manifold is toroidal, as the bump rolls off one edge it seamlessly re-emerges on the opposite edge, mathematically generating the infinitely repeating, hexagonal periodic grid fields observed in biological recording chambers.

10. Comparative Neurobiology, Behavioral Paradigms, and Lesion Deficits

10.1 Standardized Behavioral Tests of Spatial Function

The empirical dominance of the cognitive map theory was cemented through the development of specialized behavioral apparatuses designed to rigorously evaluate allocentric spatial navigation while controlling for egocentric and olfactory cues. The three most influential paradigms are:

• The Morris Water Maze: Developed by Richard Morris in 1981, this paradigm consists of a large, featureless circular pool filled with opaque, milky water. Submerged millimeters beneath the surface in a fixed coordinate sits an invisible escape platform. Because the water eliminates scent trails and provides no local tactile landmarks, an animal dropped into varying, randomized quadrants around the pool cannot rely on Taxon guidance or Route motor habits; it can only locate the hidden platform by triangulating its position relative to distal extramaze visual cues scattered around the laboratory walls. Intact animals learn within days to compute straight, direct vectors toward the platform. In a critical “probe trial,” the platform is removed; intact animals spend the vast majority of their search time swimming focused, localized search paths tightly focused within the precise quadrant where the platform previously resided.
• The Radial Arm Maze: Engineered by David Olton, this apparatus features a central circular platform from which eight (or more) identical corridors radiate like the spokes of a wheel, with food rewards placed at the ends of the arms. Optimal performance requires the animal to retrieve all food rewards without re-entering an arm that it has already depleted. The radial arm maze cleanly dissociates two distinct memory operations: Spatial Reference Memory (learning which arms are consistently baited across days, an allocentric rule) and Spatial Working Memory (tracking which arms have already been visited during the ongoing, single behavioral session).
• The T-Maze / Y-Maze Alternation: Evaluates spontaneous or rewarded spatial alternation. An animal placed in the stem must choose to enter either the left or right goal arm. On the subsequent run, intact animals exhibit an innate, ethological drive to explore the unvisited arm (spontaneous alternation), an operation that requires holding the prior spatial visit in temporary memory storage.

10.2 Lesion and Inactivation Studies across Mammalian Taxa

Application of these standardized behavioral paradigms revealed the consequences of selective hippocampal disruption. When researchers moved away from crude aspiration or electrolytic lesions—which damaged passing fiber tracts—and implemented precise excitotoxic lesions utilizing ibotenic acid or NMDA, the specificity of the cognitive map deficit was laid bare.

Animals with selective, complete bilateral lesions of the dorsal hippocampus exhibit an absolute, irrecoverable failure on the hidden platform version of the Morris water maze. They swim in aimless, random loops, failing to calculate the platform’s spatial coordinates despite weeks of continuous training. In probe trials, their search times are distributed uniformly and randomly across all four quadrants. Yet, if the exact same lesioned animals are tested on a “cued” water maze paradigm—where the escape platform is raised visibly above the water’s surface or marked with a prominent, colorful beacon—their escape latencies and swimming speeds are completely normal. The Taxon/guidance system (mediated by neocortical and striatal networks) is pristine; only the allocentric locale mapping system is destroyed.

To eliminate concerns regarding developmental or structural compensation in chronic lesion models, contemporary researchers utilize reversible pharmacological inactivation (infusions of the GABA-A agonist muscimol) or optogenetic silencing (activating inhibitory archaerhodopsin or halorhodopsin proton/chloride pumps via implanted optical fibers). Transiently silencing dorsal CA1 or CA3 specifically during the acquisition phase completely prevents the formation of the spatial map. Conversely, silencing these structures during the probe recall trial destroys the animal’s ability to retrieve an established map, with normal navigation capabilities recovering within hours as the pharmacological agent washes out or the lasers are extinguished.

10.3 Phylogenetic Diversity: Avian, Primate, and Human Mapping Mechanisms

The principles of the cognitive map theory transcend rodent neurobiology, exhibiting profound evolutionary conservation across diverse avian and mammalian taxa:

• Avian Spatial Mapping and Seasonal Neuroplasticity: Food-caching bird species, such as the black-capped chickadee (Poecile atricapillus) and Clark’s nutcracker (Nucifraga columbiana), cache tens of thousands of individual food items across square miles of rugged natural terrain during autumn, successfully retrieving these hidden caches months later during winter using allocentric spatial memory. Comparative neuroanatomical surveys demonstrate that food-caching avian species possess an avian hippocampus (homologous to the mammalian structure) that is disproportionately larger relative to brain and body mass than that of non-caching species. In autumn, when caching behavior accelerates, the avian hippocampus undergoes seasonal neurogenesis and dendritic expansion, swelling in physical volume to accommodate the computational demands of mapping thousands of spatial coordinates.
• Primate Allocentric Navigation in Virtual Reality: Early primate lesion studies yielded conflicting spatial results due to the species’ manual dexterity and reliance on binocular vision over whole-body locomotion. However, with the advent of immersive, whole-body primate virtual reality systems, electrophysiologists have recorded classical place cells, grid cells, and head-direction cells in freely viewing rhesus macaques (Macaca mulatta), operating under computational algorithms identical to the rodent architecture.
• Structural Neuroplasticity in Human Navigational Experts: In humans, the physical malleability of the cognitive map was captured in the celebrated structural MRI investigations of London taxi drivers conducted by Eleanor Maguire and colleagues (“Navigation-related structural change in the hippocampi of licensed London taxi drivers”). To acquire their operating licenses, London taxi drivers must master “The Knowledge”: memorizing the labyrinthine layout of over 25,000 streets, roundabouts, and landmarks across Greater London. Maguire’s structural MRI scans revealed that licensed London taxi drivers possess a significantly larger posterior hippocampus (the primate homolog of the rodent dorsal hippocampus) compared to age-matched controls, with the volume expansion directly correlating with the number of years spent actively navigating the city. The human cognitive map is a dynamic, structurally malleable organ that physically expands in response to decades of intensive allocentric spatial computations.

11. Synthesizing Spatial Mapping with Episodic and Declarative Memory

11.1 Relational Memory Theory and Cohen-Squire Models

For decades, cognitive neuroscience was polarized by an intense theoretical rivalry between the Spatial Cognitive Map Theory of O’Keefe and Nadel and the General Declarative / Relational Memory Theory championed by Neal Cohen, Larry Squire, and Howard Eichenbaum. While O’Keefe and Nadel maintained that the hippocampus evolved as a dedicated, domain-specific geometric mapping engine, Cohen and Eichenbaum argued that spatial navigation was merely one specific manifestation of a generalized domain-general relational processor.

The relational memory framework posits that the fundamental computational function of the hippocampus is to extract, compare, and bind relationships between arbitrary, non-spatial items (faces, odors, temporal sequences, factual associations). Proponents pointed to experiments showing that animals with hippocampal lesions failed non-spatial transitive inference tasks (learning that if A > B, and B > C, then A > C) and demonstrated impairments in contextual fear conditioning, where physical navigation was not strictly required.

In recent years, these two competing theories have largely achieved a harmonious, unified synthesis. Rather than viewing space and memory as mutually exclusive, modern neuroscience recognizes that the spatial cognitive map provides the evolutionary scaffolding for episodic memory. In nature, no animal ever experiences an event outside of space; every autobiographical memory occurs at a specific physical location, at a specific point in time. By utilizing an evolutionarily conserved geometric coordinate framework, the hippocampus provides the structural matrix within which arbitrary declarative elements—who, what, and when—are bound together. Physical space is the universal filing cabinet of mammalian episodic memory.

11.2 Time Cells and Spatiotemporal Trajectories

The definitive synthesis between spatial mapping and episodic memory was solidified by the discovery of time cells within the hippocampus. Episodic memory, as classically defined by Endel Tulving, requires not just knowing where an event occurred, but tracking the chronological sequence of events—the when.

In 2011, Christopher MacDonald, Howard Eichenbaum, and colleagues (“Grid Cells and the Hippocampus as a Cognitive Map of Time”), alongside parallel work by Eva Pastalkova and György Buzsáki, recorded from CA1 pyramidal neurons during the delay phase of memory tasks. Rodents were trained to run in place on a treadmill for a fixed 10-to-20-second interval between presenting an initial stimulus and receiving a paired reward. During this delay, the animal’s physical coordinates remained completely static; yet, individual CA1 neurons fired in robust, highly organized temporal sequences. One cell fired from seconds 0 to 2, another from seconds 2 to 4, another from seconds 4 to 7, systematically bridging the delay interval.

These time cells behave identically to place cells, except that their receptive fields are mapped across continuous time rather than physical distance. Just as place cells tile physical terrain, hippocampal time cells tile temporal intervals. This dual spatiotemporal coding mechanism provides the neural substrate for the Hippocampal Indexing Theory, originally formulated by Robert Teyler and Jerry DiScenna. The hippocampus acts as an index of cortical coordinates: when an episodic memory is encoded, the hippocampus registers a unique spatiotemporal coordinate vector that points to the distributed neocortical sensory assemblies processing the sights, sounds, and emotions of the event. To retrieve an episodic memory is to mentally navigate backward along this spatiotemporal trajectory, reactivating the hippocampal index, which in turn reconstructs the full subjective experience across the neocortex.

11.3 Neural Replay, Sharp-Wave Ripples, and Memory Consolidation

The final operational phase of the hippocampal cognitive map occurs during offline states—periods of quiet wakefulness, consummatory behavior, and non-Rapid Eye Movement (NREM) slow-wave sleep. During these states, the regular 8 Hz theta rhythm vanishes, replaced by transient, high-amplitude voltage deflections in the local field potential known as sharp waves, which originate in CA3 and trigger high-frequency (150–250 Hz) oscillatory events in CA1 designated as ripples (collectively, sharp-wave ripples, or SWRs).

In a series of landmark discoveries initiated by Matthew Wilson and Bruce McNaughton in 1994, electrophysiologists discovered the phenomenon of neural replay. During a sharp-wave ripple, the specific ensembles of place cells that fired sequentially as an animal navigated a physical path during prior wakefulness suddenly fire again, compressed temporally by a factor of 10 to 20 times (unfolding within 50 to 100 milliseconds). Replay occurs in two distinct operational modes:

• Forward Replay: Replays the spatial trajectory in the exact temporal order it was experienced. During wakeful contemplation at a choice point, forward replay often simulates future, unvisited trajectories, acting as an internal simulation of potential paths to guide decision making.
• Reverse Replay: Replays the trajectory in reverse chronological order, starting from the reward location and propagating backward to the origin point. This backward sweep is functionally critical for reinforcement learning, propagating dopamine-mediated reward signals back along the path to bind the behavioral sequence via STDP.

Sharp-wave ripples and neural replay represent the physical dialogue of two-stage systems memory consolidation. The high-frequency SWRs propagate out of the subiculum and entorhinal cortex, traveling via long-range reciprocal projections to the medial prefrontal cortex and distributed neocortical association areas. By repeatedly bombarding the neocortex with compressed spatiotemporal replays of daytime experiences, the hippocampus gradually transfers the memory trace to the neocortex. Over time, the neocortical connectivity becomes structurally self-sustaining, liberating the memory from absolute dependence on the hippocampal index and synthesizing individual spatial episodes into generalized semantic schemas.

12. Contemporary Advances, Computational Paradigms, and the Theory’s Legacy

12.1 Modern Optogenetic and Multi-Photon Calcium Imaging Horizons

The study of hippocampal cognitive mapping has undergone an unprecedented renaissance driven by cutting-edge optical and electrophysiological innovations. Chief among these is in vivo two-photon and multi-photon calcium imaging combined with miniaturized, head-mounted fluorescent microscopes (miniscopes), pioneered by Mark Schnitzer and David Tank. By genetically encoding fluorescent calcium indicators (such as the GCaMP family) into specific hippocampal subfields, researchers can visualize the simultaneous, real-time activity of thousands of individual, identified place cells across months of exploration.

These optical imaging platforms have solved long-standing questions regarding the chronic stability and representational drift of the cognitive map. Long-term tracking reveals that while a stable “core” ensemble of place cells preserves the spatial representation across an animal’s lifespan, a fraction of the ensemble undergoes continuous, slow representational drift over weeks. This balance provides the hippocampus with computational flexibility: preserving spatial orientation via the core ensemble while encoding novel contextual and temporal variations across the drifting subpopulation.

Concurrently, the development of high-density Neuropixels silicon probes has allowed the simultaneous recording of thousands of isolated single units across distributed networks—spanning the dentate gyrus, CA3, CA1, the subicular complex, the medial and lateral entorhinal cortices, and the retrosplenial cortex in a single animal. Researchers can now observe the step-by-step transformation of grid codes into place fields and monitor the multi-regional coordination of replay during sharp-wave ripples at single-spike millisecond resolution. Furthermore, with all-optical interrogation—combining two-photon imaging with targeted, single-cell optogenetic photo-stimulation—investigators can artificially reactivate specific, chosen place cell ensembles in an awake animal, effectively implanting synthetic spatial memories and steering the animal’s navigational choices across virtual environments.

12.2 Deep Learning and Computational Models of Cognitive Mapping

The theoretical framework of O’Keefe and Nadel has exerted a transformative influence on modern artificial intelligence, deep learning, and computational neuroscience. A central question in AI has been how an artificial agent can navigate complex, unstructured environments without being fed continuous, hand-crafted cartographic data.

In a groundbreaking 2018 study published by Andrea Banino and colleagues at Google DeepMind (“Vector-based navigation using grid-like representations in artificial agents”), deep recurrent neural networks (RNNs) equipped with long short-term memory (LSTM) architectures were trained to perform path integration using raw velocity inputs in simulated environments. Remarkably, without any structural guidance or pre-programmed constraints imposed by the researchers, the artificial units in the intermediate layers spontaneously organized themselves into periodic hexagonal grid-like firing fields that precisely mirrored biological grid cells in the mammalian entorhinal cortex. This emergent solution proved that grid-like representations are the mathematically optimal computational solution for path integration in 2D space. When these artificial grid representations were integrated into downstream reinforcement learning networks, the agents exhibited superhuman navigation, calculating complex, novel shortcuts across labyrinthine virtual mazes.

Simultaneously, computational neuroscientists have developed models of cognitive mapping based on the Successor Representation (SR), originally formulated in reinforcement learning by Peter Dayan and applied to the hippocampus by Kimberly Stachenfeld, Matthew Botvinick, and Samuel Gershman. The successor representation model posits that place cells do not encode absolute physical coordinates ($x, y$), but rather the discounted probability of occupying future states along an animal’s expected transition policy:

$$M(s, s’) = \mathbb{E}\left[\sum_{t=0}^{\infty} \gamma^t \mathbb{I}(s_t = s’) mid s_0 = s\right]$$

By defining the cognitive map as a predictive matrix of future transitions, the successor representation model elegantly explains why place fields cluster near rewards, stretch along preferred behavioral trajectories, and alter their firing shapes in response to environmental barriers, synthesizing metric spatial mapping with predictive reinforcement learning algorithms.

12.3 Clinical Implications: Neurodegeneration, Aging, and Spatial Pathology

Beyond theoretical modeling and laboratory neurobiology, the cognitive map theory provides vital clinical insights into neuropathology, particularly the diagnosis and progression of Alzheimer’s disease (AD). Pathologically, Alzheimer’s disease does not target the brain uniformly; its earliest molecular lesions—neurofibrillary tangles composed of hyperphosphorylated tau protein—manifest with selective vulnerability within the transentorhinal and medial entorhinal cortices (Braak Stage I and II) before spreading into the CA1 subfield and hippocampus proper.

Because the medial entorhinal cortex and hippocampus harbor the foundational neural engines of the cognitive map (grid cells and place cells), the very first behavioral manifestation of preclinical Alzheimer’s disease is rarely generic linguistic memory loss; it is spatial disorientation and the collapse of allocentric navigation. Patients lose the ability to maintain an internal map of familiar neighborhoods, repeatedly getting lost, struggling to estimate distances, and failing to compute novel detour routes. Traditional cognitive paper-and-pencil assessments (such as the Mini-Mental State Examination) are notoriously insensitive to these subtle metric navigational disruptions, often failing to detect the disease until extensive neocortical damage has already occurred.

Armed with the foundational insights of the O’Keefe-Nadel framework, translational clinical neuroscientists—such as Dennis Chan, Michael Hornberger, and Gillian Coughlan—have pioneered immersive virtual reality (VR) spatial navigation diagnostics. These digital biomarkers, exemplified by spatial tasks such as the “Supermarket Game” and the “Sea Hero Quest” platform, test an individual’s specific capacity to perform path integration, boundary-referenced orientation, and allocentric vector navigation. These non-invasive VR navigation assessments have demonstrated high diagnostic accuracy in identifying cognitively asymptomatic individuals harboring elevated cerebrospinal fluid (CSF) tau and amyloid biomarkers, decades before the onset of overt dementia. The conceptual framework first drafted by John O’Keefe and Lynn Nadel in 1978 has thus evolved into a frontline clinical tool for early detection and therapeutic intervention in human neurodegenerative disease.

Conclusion

The publication of The Hippocampus as a Cognitive Map in 1978 stands as a defining watershed in the history of neuroscience. John O’Keefe and Lynn Nadel achieved what few scientific treatises ever accomplish: an enduring, paradigm-shifting synthesis that successfully integrated the philosophical inquiries of Immanuel Kant, the ethological and behavioral psychology of Edward Tolman, the cellular precision of single-unit electrophysiology, and the structural beauty of hippocampal microcircuitry. By demonstrating that the dorsal hippocampus constructs an internal, allocentric representation of absolute space, they overturned the reductive behavioral orthodoxy of stimulus-response habit chains and placed internal representations squarely at the center of modern cognitive neurobiology.

The subsequent decades of research have not diminished the core tenets of their thesis; rather, they have broadened and enriched them. The discoveries of head direction cells, grid cells, border cells, time cells, and neural replay have completed the intricate computational architecture of the extended hippocampal-entorhinal mapping system. Today, the cognitive map theory is recognized not merely as a model of physical terrain navigation, but as the fundamental evolutionary scaffold upon which the mammalian brain constructs episodic memories, plans prospective futures, organizes semantic concepts, and traverses the abstract multidimensional landscapes of human cognition. Through its profound theoretical coherence, empirical robustness, and clinical relevance, O’Keefe and Nadel’s cognitive map theory remains an intellectual monument—a testament to the power of unifying cellular physiology with the deepest questions of the conscious mind.

References

• Aronov, D., Nevers, R., & Tank, D. W. (2017). Mapping of a non-spatial dimension by the hippocampal-entorhinal circuit. Nature, 543(7647), 719–724. https://doi.org/10.1038/nature21692
• Banino, A., Barry, C., Uria, B., Blundell, C., Lillicrap, T., Mirowski, P., Pritzel, A., Chadwick, M. J., Degris, T., Modayil, J., Wayne, G., Soyer, H., Viola, F., Zhang, B., Goroshin, R., Evans, N. C., Stachenfeld, K., Hassabis, D., & Kumaran, D. (2018). Vector-based navigation using grid-like representations in artificial agents. Nature, 557(7705), 429–433. https://doi.org/10.1038/s41586-018-0102-6
• Bliss, T. V., & Lømo, T. (1973). Long-lasting potentiation of synaptic transmission in the perforant path of the anaesthetized rabbit to stimulation of the perforant path. The Journal of Physiology, 232(2), 331–356. https://doi.org/10.1113/jphysiol.1973.sp010273
• Burgess, N., Jackson, A., Hartley, T., & O’Keefe, J. (2000). Predictions and tests of a model of spatial memory using grid and place cells. Philosophical Transactions of the Royal Society of London. Series B: Biological Sciences, 355(1404), 1699–1711. https://doi.org/10.1098/rstb.2000.0734
• Buzsáki, G. (2015). Hippocampal sharp wave-ripple: A cognitive biomarker for episodic memory and planning. Hippocampus, 25(10), 1073–1188. https://doi.org/10.1002/hipo.22488
• Constantinescu, A. O., O’Reilly, J. X., & Behrens, T. E. (2016). Organizing conceptual knowledge in humans with a gridlike code. Science, 352(6292), 1464–1468. https://doi.org/10.1126/science.aaf0941
• Eichenbaum, H. (2000). A cortical-hippocampal system for declarative memory. Nature Reviews Neuroscience, 1(1), 41–50. https://doi.org/10.1038/35036213
• 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
• Lever, C., Burton, S., Jeewajee, A., O’Keefe, J., & Burgess, N. (2009). Boundary vector cells in the subiculum of the hippocampal formation. The Journal of Neuroscience, 29(31), 9771–9777. https://doi.org/10.1523/JNEUROSCI.1319-09.2009
• MacDonald, C. J., Lepage, K. Q., Eden, U. T., & Eichenbaum, H. (2011). Grid cells and the hippocampus as a cognitive map of time. Neuron, 71(4), 737–749. https://doi.org/10.1016/j.neuron.2011.07.012
• Maguire, E. A., Gadian, D. G., Johnsrude, I. S., Good, C. D., Ashburner, J., Frackowiak, R. S., & Frith, C. D. (2000). Navigation-related structural change in the hippocampi of licensed London taxi drivers. Proceedings of the National Academy of Sciences, 97(8), 4398–4403. https://doi.org/10.1073/pnas.070039597
• Marr, D. (1971). Simple memory: A theory for archicortex. Philosophical Transactions of the Royal Society of London. Series B, Biological Sciences, 262(841), 23–81. https://doi.org/10.1098/rstb.1971.0078
• Morris, R. G. (1984). Developments of a water-maze procedure for studying spatial learning in the rat. Journal of Neuroscience Methods, 11(1), 47–60. https://doi.org/10.1016/0165-0270(84)90007-4
• 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: Clarendon Press. https://www.cognitivemap.net/
• O’Keefe, J., & Recce, M. L. (1993). Phase relationship between hippocampal place units and the EEG theta rhythm. Hippocampus, 3(3), 317–330. https://doi.org/10.1002/hipo.450030307
• Olton, D. S., & Samuelson, R. J. (1976). Remembrance of places passed: Spatial memory in rats. Journal of Experimental Psychology: Animal Behavior Processes, 2(2), 97–116. https://doi.org/10.1037/0097-7403.2.2.97
• Pastalkova, E., Itskov, V., Amarasingham, A., & Buzsáki, G. (2008). Internally generated cell assembly sequences in the rat hippocampus. Science, 321(5894), 1322–1327. https://doi.org/10.1126/science.1159775
• Quiroga, R. Q., Reddy, L., Kreiman, G., Koch, C., & Fried, I. (2005). Invariant visual representation by single neurons in the human brain. Nature, 435(7045), 1102–1107. https://doi.org/10.1038/nature03687
• Scoville, W. B., & Milner, B. (1957). Loss of recent memory after bilateral hippocampal lesions. Journal of Neurology, Neurosurgery, and Psychiatry, 20(1), 11–21. https://doi.org/10.1136/jnnp.20.1.11
• Solstad, T., Boccara, C. N., Kropff, E., Moser, M. B., & Moser, E. I. (2008). Representation of geometric borders in the entorhinal cortex. Science, 322(5909), 1865–1868. https://doi.org/10.1126/science.1166466
• Stachenfeld, K. L., Botvinick, M. M., & Gershman, S. J. (2017). The hippocampus as a predictive map. Nature Neuroscience, 20(11), 1643–1653. https://doi.org/10.1038/nn.4650
• 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. The Journal of Neuroscience, 10(2), 420–435. https://doi.org/10.1523/JNEUROSCI.10-02-00420.1990
• Teyler, T. J., & DiScenna, P. (1986). The hippocampal memory indexing theory. Behavioral Neuroscience, 100(2), 147–154. https://doi.org/10.1037/0735-7044.100.2.147
• Tolman, E. C. (1948). Cognitive maps in rats and men. Psychological Review, 55(4), 189–208. https://doi.org/10.1037/h0061626
• Wilson, M. A., & McNaughton, B. L. (1994). Reactivation of hippocampal ensemble memories during sleep. Science, 265(5172), 676–679. https://doi.org/10.1126/science.8036517