In 1971, John O'Keefe and Jonathan Dostrovsky lowered a fine wire electrode into a rat's hippocampus and watched a single neuron fire almost exclusively when the animal stood in one corner of a small testing box. The neuron did not care much about what the rat was doing there. It cared about where the rat was. That result, published in the journal Brain Research, became the founding observation of place cells, and it set a research program that has shaped how neuroscientists think about space, memory, and the brain's internal map for more than five decades.

Place Cells and Their Firing Fields

The unit O'Keefe described was the place field: a region of the environment where a given pyramidal neuron reliably fires. Each cell has its own field, and the fields of many cells tile an enclosure so that the animal's position can in principle be read out from which cells are active. The elegance was hard to resist. A continuous, messy environment had been reduced to a discrete, stable unit that a physiologist could count.

What made the finding durable was its indifference to behavior. A rat could be walking, grooming, or paused at a food cup, and the same cell fired when it crossed the same patch of floor. O'Keefe and Lynn Nadel argued in their 1978 book The Hippocampus as a Cognitive Map that this population of cells formed an internal representation of space, a map the animal consulted rather than a reflex it executed. The map was the point.

Most recordings happened in small, uniform enclosures, often a square box or a radial arm maze, with the room around it held as constant as possible. Position and firing rate were the two variables that mattered. Everything else—wall color, bedding scent, time of day—was noise to be controlled away. That choice was methodologically sound and quietly consequential.

The Push to Standardize Spatial Mapping

Three decades later, Edvard Moser and May-Britt Moser, working in Trondheim, recorded from the entorhinal cortex, one synapse upstream of the hippocampus. In 2005 they reported cells that fired in a striking hexagonal lattice across an open arena, a pattern that repeated at regular intervals as the animal moved. These were grid cells, and they looked like a coordinate system: periodic, metric, and independent of any particular landmark.

The grid discovery gave the field a tidy hierarchy. Grid cells in entorhinal cortex were proposed to provide a path-integration signal, a running estimate of position based on self-motion. Place cells in the hippocampus were described as reading out combinations of grid inputs, sharpening a periodic code into a single location. Space became a computational pipeline with an input layer and an output layer.

That model traveled well. Grid-like signals were later reported in humans using functional imaging, and the 2014 Nobel Prize in Physiology or Medicine went to O'Keefe and the Mosers for the work. The framework had become the standard reference for how the brain represents location. It also made the environment itself increasingly abstract: what mattered was geometry and distance, not the specific room an animal happened to inhabit.

What the Mapping Framework Left Out

Place fields are not as fixed as the textbook picture suggests. Change the shape of the enclosure, or rotate a prominent visual cue, and many place cells remap: they stop firing where they fired before and start firing somewhere else. This phenomenon, called remapping, was documented early and has been studied intensively. It is the clearest evidence that the map is not a static coordinate grid but something the brain rebuilds in response to its surroundings.

Remapping sits awkwardly with the clean hierarchy. If place fields are derived from grid inputs, why should a new wall color reorganize them so completely? The field's working answer has been to invoke context, a term that does a lot of unexplained work. Context is whatever changed when the room changed, and it is usually treated as a switch that flips the map rather than as a rich signal in its own right.

Other cell types complicate the picture further. Boundary cells fire near walls and edges. Head-direction cells track which way the animal faces. Object-vector cells respond to the location of specific items. These cells are well documented, and they are often described as inputs to the spatial system rather than as parts of the environment being represented. The room's contents are present in the data and absent from the story.

Consequences for Memory and Cognition

O'Keefe and Nadel's cognitive map was explicitly spatial. Memory entered the picture because the hippocampus also seemed necessary for forming new episodic memories, and the book proposed that episodes are stored as routes through a map. Navigation became the master metaphor for remembering: to recall an event is to retrace a path. The metaphor is powerful and has been productive, but it makes location the organizing principle of memory by assumption.

Subsequent work has shown that hippocampal neurons carry information well beyond location. In tasks where rats run the same maze for different rewards, many place cells fire more strongly on trials leading to a preferred outcome, even when the animal's trajectory is identical. The firing rate changes with value, not with position. Similarly, during sharp-wave ripples in sleep and quiet rest, the hippocampus reactivates sequences of place cells that correspond to recently traveled routes, but the sequences are often replayed in reverse and at variable speeds, suggesting they serve a function beyond simple path rehearsal. These findings indicate that location is one dimension among several that the hippocampus encodes, and that the spatial framework alone does not predict how value or reward expectation will alter firing.

Neuroscientists have started to ask what else the hippocampus encodes when the task is not a maze. Odor cues, reward locations, and abstract variables like elapsed time all modulate hippocampal firing. The spatial framework did not predict this, and fitting the results into it requires stretching the definition of location until it covers almost anything.

Reintroducing the Room

A growing body of work treats the environment as a variable to be measured rather than a constant to be controlled. Experiments that systematically vary multimodal cues, including visual landmarks, odors, textures, and enclosure shape, find that place fields respond to all of them, with different cells weighting different cues. The map is not purely geometric. It is anchored to whatever the animal can perceive.

Recording during free exploration in larger, more naturalistic environments has also changed the picture. In open fields and multi-compartment arenas, place fields are less stable and more context-dependent than in the standard square box. Some of what looked like a fixed map may have been an artifact of a very repetitive room. The finding does not overturn the place cell literature, but it does suggest that the field's canonical preparation was unusually constraining.

Treating context as a latent variable, something to be inferred from behavior and neural activity rather than assumed, is one way forward. This framing keeps the spatial results and adds a second question: what else is the hippocampus tracking, and how does that interact with location? The question is old. The tools to answer it are newer.

Practical Steps for Researchers

Vary room features deliberately across trials, including wall patterns, odors, and floor texture, and report the full cue configuration in the methods rather than describing the arena as a plain box. A reader cannot interpret remapping without knowing what changed.

Track behavior beyond position. Speed, heading, grooming, and reward approach all modulate hippocampal firing, and recording them costs little compared with the electrode implant itself.

• Log the cue configuration for every session, not just the first.
• Record grid and place cells in the same animals where the preparation allows it.
• Analyze remapping as an encoding event, not only as a failure of stability.
• Test at least one naturalistic enclosure alongside the standard small box.

Model context as a latent variable with observable causes, so that the model can be checked against the room. A latent variable that never touches a measured cue is a placeholder, and placeholders accumulate. The grid cell literature is precise enough to survive this scrutiny. The context term, so far, is not.