In mammals, including humans, a dedicated neural network processes spatial information. Place cells in the hippocampus fire only when an animal reaches a particular spot, and their collective activity forms an internal “cognitive map.” Grid cells in the medial entorhinal cortex fire at hexagonally arranged points, providing a metric for distance and direction. Together with head-direction and border cells they underpin memory, learning, and even inspire virtual-reality systems and autonomous robots. Because this pathway is among the first damaged in Alzheimer’s disease, it offers crucial clues for early diagnosis and therapy.

Place cells in hippocampal CA1/CA3 exhibit location-specific spiking that cannot be explained by simple sensory cues; O’Keefe framed them as the neural substrate of the “cognitive map.” Decades later, the Mosers recorded from the medial entorhinal cortex (MEC), the main cortical input to the hippocampus, and identified grid cells whose multiple firing fields form a regular hexagonal lattice. Grid spacing increases exponentially along the dorsal–ventral MEC axis, implying a modular organization. Continuous attractor and oscillatory interference models account for these patterns, supported by observations of theta phase precession and context-dependent remapping. Bidirectional synaptic loops link grid and place networks, and amyloid pathology disrupts their plasticity and spatial coding in transgenic mice. Beyond navigation, this metric code is debated as a scaffold for episodic memory and even abstract concept representation, bridging systems neuroscience with computational models and artificial intelligence.

Unit recordings initiated by O’Keefe & Dostrovsky (1971) revealed spatially selective firing fields in CA1 pyramidal cells and subsequent experiments characterized experience-dependent remapping. The Mosers identified MEC layer II/III neurons with periodic, phase-coherent hexagonal firing—grid cells—whose spacing λ discretizes from ~30 cm dorsally to >3 m ventrally (Stensola et al. 2012). Continuous attractor models using symmetric recurrent connectivity and velocity-modulated input explain stable grid translations, while vestibular inactivation and theta disruption provide in vivo support. Optogenetic silencing indicates that retrograde CA1→MEC projections are required to sustain grid amplitude, implying re-entrant hippocampo-entorhinal loops that integrate theta phase precession and sequence replay. Intracranial recordings and fMRI detect hexadirectional BOLD modulation in humans, validating translational relevance. In Tg2576 Alzheimer mice, Aβ-associated MEC-CA1 LTP deficits coincide with place-field dispersion, highlighting molecular substrates of navigational deficits. Future directions include grid coding in high-dimensional tasks and its implementation in deep reinforcement-learning architectures.