Anyone who has walked home from work along a familiar route without consciously choosing each turn has experienced something of what neuroscientists call a cognitive map. The term, coined by the American psychologist Edward Tolman in 1948, refers to the internal mental model that allows an animal - or a person - to navigate an environment flexibly, without relying on a rigid sequence of memorised turns. In the decades since Tolman's original experiments with rats in mazes, researchers have steadily accumulated evidence that a cognitive map is not a metaphor but a real neural structure with identifiable components.

The key discovery came in 1971, when the neuroscientist John O'Keefe recorded the activity of individual neurons in the hippocampus of rats as they explored a simple box-shaped environment. O'Keefe found that each neuron fired only when the rat was in a particular location. Move the rat to a different part of the box and a different neuron would take over. He called these cells place cells, and the name has stuck. Over the next two decades, recordings from other hippocampal regions revealed grid cells, which fire at regular geometric intervals across the entire environment, and head-direction cells, which fire when the animal's head is pointed in a particular compass direction. Together, these cell types provide the raw material of a neural map: position, orientation, and the geometric relationships between places.

The discoveries were not confined to rats. Similar cells have since been identified in bats, monkeys, and, through studies of patients undergoing surgery for epilepsy, in human beings. The human version is somewhat more flexible than that of other animals. People produce place cells not only for physical locations but for positions in abstract spaces: the neural recording of a volunteer thinking about a sequence of family members, or navigating the arrangement of objects on a table, shows many of the same patterns seen when the volunteer is walking around a room. This observation has led the Israeli neuroscientist Yonatan Rosenberg to argue that the cognitive map is a general-purpose system, used by the brain for organising any kind of related information, not just physical environments.

Navigation itself is only one of the functions the system performs. Memory formation seems to rely heavily on the same neural structures. A patient who has lost the use of the hippocampus - through surgery, illness, or an accident - typically has severe difficulty forming new memories of events, a condition known as anterograde amnesia. Such patients can often remember childhood and hold a conversation, but they cannot recall what they did an hour ago. The long-standing explanation is that memories of events are tagged with the place in which they occurred and that the place tag is stored in the hippocampus; losing the tagging system leaves the memories themselves inaccessible.

The cognitive map is not flawless. People who have been brought up in environments with few long straight lines, such as traditional island fishermen, are typically better than city residents at estimating distance over water. City residents, however, are better at recognising a place from an unusual angle, an ability which seems to depend on the repeated experience of walking past the same building from different sides. Both groups show equally strong basic map-like neural activity, but the details of what they have mapped differ systematically. Rosenberg has described the cognitive map as a "trained muscle", shaped by the specific environments a person has encountered.

Aging affects the system in predictable ways. Grid cells, in particular, appear to become less precise after the age of about sixty-five, which may explain the common observation that older people find it harder to orient themselves in unfamiliar places. The effect is not merely anecdotal: hospital studies have shown that patients with early Alzheimer's disease perform particularly poorly on tasks that require navigation in a virtual environment, even before they show obvious problems with ordinary memory. Some researchers have argued that a carefully designed navigation test could become an early diagnostic for Alzheimer's, detecting the disease years before the current clinical criteria would recognise it.

The practical implications of this research are already reaching beyond medicine. Designers of public spaces have begun to pay attention to what they call legibility: the ease with which a visitor can build an accurate cognitive map of a building or a neighbourhood. Hospitals with poor legibility generate large numbers of missed appointments and stressed visitors, and the architect who builds a well-signposted, well-proportioned corridor is not only serving aesthetic goals but making a measurable contribution to the mental workload of every user. Urban planners speak of landmarks and of the importance of distinctive corners and changes of level, features that the brain's place cells evidently find easier to encode than a uniform grid of indistinguishable streets.

Technology, meanwhile, has introduced a new variable into the system. Reliance on satellite navigation appears to reduce the activity of the hippocampus during route finding, and several recent studies have suggested that heavy users of navigation apps are worse at finding unfamiliar destinations without their phones than people who use maps or rely on memory. The effect is not yet well quantified, and it may reverse when the app is switched off. Rosenberg has speculated, however, that a cognitive map is a perishable asset: if the brain is not required to build and maintain one, the relevant circuits may not be rehearsed, and the ability to navigate complex environments from memory may quietly weaken.