Brain’s Navigational System

The ability to know one’s location in the environment and navigate to important goals is important for the survival of humans, mammals, and many mobile species. Over the past decades, research in the area of spatial cognition has been highly productive, drawing interest from researchers as far afield as psychology, neuroscience, and mathematics. This research has produced insight into some of the strategies used by animals during navigation and, importantly, identified a number of cell types specialized for processing spatial information, providing a framework for understanding the neural representations and mechanisms underlying this fundamental cognitive ability. Some of the core findings from this field are summarized below.

Cognitive Map

Behavioral theory dominated psychology in the early decades of the 20^th century. Behaviorism, in short, asserts animal behavior is the simple result of learned stimulus-response sequences that lead to a desirable outcome. Most behavioral research at that time (and today) was carried out on rats trained to navigate through mazes of varying complexity in order to receive a food reward. According to behaviorists, rats learn to navigate by learning a set of actions (such as a sequence of turns) that lead to their tasty reward. However, Edward Tolman was the first to challenge this view. Tolman instead postulated animals possess a set of internal representations that can flexibly guide behavior to important goals. In terms of navigation, Tolman hypothesized animals possess a so-called ‘cognitive map’ – a kind of mental image of the environment preserving information about the location of different cues and landmarks and their relation to each other that supports navigation through a complex and changing environments. Tolman tested his hypothesis with a set of experiments. In one of his experiments, rats were first trained to run down a sequence of circuitous corridors to reach their reward – a task that could be solved by a simple stimulus-response strategy. After training on this version of the task, the circuitous portion of the maze was removed, and a set of corridors organized in a ’sun-burst’ manner were inserted instead. Only one of these new corridors led to the reward. If the animals had been using a stimulus-response strategy to solve the task previously, they would not be able to solve this version of the task. However, many of them took the correct corridor – a path they had never taken before, leading them to their reward. Tolman argued the animals were able to solve this task as they had developed a mental (‘cognitive’) map of the maze environment, allowing them to take novel and flexible routes.

The Brain Simulates Paths to a Brighter Future Hippocampal place cells, which create an internal map during novel exploration, are getting ‘replayed’ during sleep

The idea animals possess internal representations that influence navigational behavior is now widely accepted. Moreover, research has identified some of the strategies animals use to guide navigation, such as landmarks and the geometric shape of the environment. To give an example, if animals are trained to navigate to a reward in a simple cylindrical environment, with the only landmark cue being a white card placed on one of its walls, if the cue card is then rotated to another location, the animal’s route to its goal will correspondingly rotate as well. Moreover, animals have also been found to rely on a process termed path integration, or dead reckoning. That is, animals are able to move between two locations in the absence of environmental information (such as in the dark), as they can integrate their ‘internal’ signals. Examples of such signals are those coming from the vestibular system detecting body movements, those from proprioception indicating limb position, and motor efference signals indicating recently commanded and executed movements. Mittelsteadt and Mittelsteadt empirically tested whether gerbils relied on path integration in a task where the rodents had to retrieve a pup from a circular arena and return to their home nest at the arena border. If the gerbils were rotated before navigating home so slowly, their vestibular system did not detect the rotation; they made an error in their homeward journey proportionate to the amount they had been rotated by, implying the gerbils were relying on their internal (vestibular) signals to solve the task.

To sum up, research over the past century has highlighted animals’ ability to rely on internal representations to guide spatial behavior and pinpointed the central strategies used by animals in goal-directed navigation. More recently, some of the biological processes supporting this ability have been identified.

Neurobiology of Self Location

Place Cells

During the early 1970s, John O’Keefe and his colleagues carried out a host of experiments on freely moving rats while concurrently recording extracellularly from a brain region in the medial temporal lobe called the hippocampus. O’Keefe and colleagues found the activity of the principal cells of the CA1 and CA3 subfields was strongly predicted by the spatial location of the animals. As such O’Keefe et al. named these cells place cells. Place cells are typically silent but robustly increase their firing when animals pass through a region of the environment where a cell’s firing field (i.e., ‘place’ field) is located. Different place cells have place fields in different parts of the environment such that at any particular spatial location, only a small number of place cells are active, providing a precise code for self-location. Moreover, on a population level, place cells provide a kind of ‘map’ of an environment, perhaps akin to Tolman’s cognitive map. For a single environment, a place cell’s firing field is stable across time, permitting the environment’s shape and dominant landmarks not to change. However, in a different environment, a place cell may change the location of its firing field or cease firing together, a process known as ‘remapping’. Thus, each environment will have a distinct place cell representation. Place cells have been intensely studied for over four decades and have been identified in different species, including mice, bats, and humans.

Interestingly, the activity of place cells seems influenced by the same variables known to guide animal navigation. Namely, distal cues (i.e., landmarks) and the geometry of the environment. For example, if animals are placed in a simple environment with the only orienting landmark being a cue card hung on one of its walls, if the cue card is then rotated by a certain amount, the location of a place cell’s place field will rotate by the same amount. Finally, place cells also seem to be able to rely on information derived from path integration. In fact, place cells have been found to prefer the use of internal signals in experimental conditions that put environmental and internal signals into conflict. However, under normal circumstances, the activity of place cells is mostly influenced by environmental information.

Head Direction Cells

Fast forward a decade later, another spatial cell type, the head direction cell, was discovered by James Ranck and colleagues. Different to place cells, head direction cells can fire action potentials anywhere in the environment. However, a head-direction cell only becomes active when the animal’s head is oriented in the cell’s preferred direction in the horizontal (azimuth) plane. Each cell has a different preferred direction, and thus, collectively, these cells may underpin one’s sense of direction’. Head direction cells were first discovered in the presubiculum but have since been identified in a wide array of brain regions, both cortical as well as sub-cortical, such as thalamic nuclei, the mammillary bodies, and the entorhinal cortex, some of which project directly to the hippocampus.

Internal metric of space is influenced by geometry of environment New Nature paper reveals that brain perceives distance differently depending on surroundings

Similar to place cells, head direction cells rely on environmental and self-motion cues. Conditions that lead to place cell re-mapping lead to a concomitant rotation of head direction cells. However, head direction cells differ from placing cells in that they are active in all environments, and when they rotate, they do so coherently as a population. For example, if in a given environment, one cell has a preferred direction at 60° and another at 120°, when the animal moves to a different environment the two cells will rotate their preferred firing direction together to maintain the same angular offset of 60°.

Through the 1980s and 1990s, the field of neuroscience dedicated to understanding the neural representations underlying spatial cognition was a highly productive one. However, it had to wait two decades for another major discovery.

Grid Cells

In 2005, the active field of hippocampal research was re-invigorated by the discovery made by May-Britt Moser and Edvard Moser of yet another type of cell that seemed dedicated to processing spatial information. Similar to place cells, this cell type has spatial firing fields in an environment. Yet, rather than just having one firing field per environment, it has multiple fields tessellating entire environments in a remarkably regular triangular pattern. Due to their regular and repetitive nature, the Mosers termed these cells grid cells. Grid cells are thought to be most numerous in the superficial layers of the medial entorhinal cortex (MEC) but are also found in the deeper layers of the MEC as well as in the pre-and parasubiculum, where they are often also modulated by head direction.

A grid cell can be described in three ways; in terms of its scale (distance between adjacent firing fields), orientation (of its grid axes to some reference direction), and phase (the two-dimensional offset of the grid axes to an external reference point). Moreover, grid cells are anatomically organized into modules that share scale and orientation but whose phases are offset by different amounts. A grid cell’s phase and orientation may shift when an animal goes into different environments, but similar to head direction cells, it is active in all environments. Importantly, shifts in the orientation and phases of grid cells are coherent within a module but may vary between modules.

When first discovered, grid cells were proposed as the neural substrate for path integration, given their implicit coding for direction and distance and their invariant activity across different environments. Moreover, their discovery attracted the attention of many theoreticians who speculated that grid cells could play an important role in shaping the activity of other spatial cells, especially place cells. However, research in recent years has cast doubt on these speculations. Namely, place cell activity persists in the absence of grid cell activity, and mounting evidence indicates grid cells are also influenced by environmental information such as the geometry and novelty of the environment. However, scientists still contend grid cells are the most likely substrate for path integration and undoubtedly influence the activity of place cells, although they may not determine it.

Boundary Cells

As mentioned at various points above, the geometry of the environment seems to exert a strong influence over the activity of the different spatial cell types. In fact, early models of place cell firing assumed the existence of ‘boundary’ cells that code for distance to the nearest environmental border and provide input to place cells. The prediction of this model was that these cells should have elongated fields along a particular environmental border and be controlled by a single-head direction. Such cells were indeed identified independently by the Mosers and O’Keefe research groups just under a decade ago in the subiculum and the medial entorhinal cortex of rats. As predicted, the cells fire close to environmental boundaries, such as walls or sharp drop edges, and are strongly modulated by head direction. To give an example, a boundary cell may be strongly tuned to a boundary that lies to the south of the animal. As such, if a second boundary is inserted, running parallel to the first boundary, the cell will develop a second firing field along the north edge of the new boundary. In terms of the influence of boundary cells over place cell activity, this question remains to be tested directly, yet it is now known boundary cells project to the hippocampus. Therefore, it is conceivable their activity shapes that of place cells.

In sum, a handful of cells that seem specialized for processing spatial information have been identified in rodent and mammalian brains. These cells seem well posited to support an animal’s wayfinding abilities. However, how does this all relate to humans?

Implications for Human Spatial Cognition

It is well established that the human hippocampus is required for memory formation and, at least, short-term memory storage and retrieval. However, until recently, it had not been clear whether the human hippocampus, similar to that of rodents, was important for spatial behavior. Recent technological progress made in the field of brain imaging (e.g., functional MRI) has afforded neuroscientists the opportunity to study the involvement of different regions of the human brain in cognitive processes. In agreement with the aforementioned findings from animal experiments, this research has revealed the involvement of the hippocampus in various spatial tasks, especially tasks that require flexible way-finding strategies. Moreover, patients with intracranial electrodes – implanted for clinical purposes, have been provided unique opportunities to study the single-neuron activity underlying human cognition. A handful of these studies have been carried out using spatial tasks and have identified cells in the medial temporal lobe with similar response properties to place and grid cells found in rodents. Furthermore, the discoveries made through animal experiments have inspired research on patients with damage to the hippocampus. These studies have found hippocampal patients generally show deficits in spatial behavior, such as goal-directed navigation. Finally, common amnesic conditions, such as Alzheimer’s disease, are now known to also be associated with impairments in spatial cognition. Together, these findings imply the neural representations that underpin spatial cognition in rodents are similar to those in humans.

In sum, although the majority of research into the brain’s navigation system has been carried out on rodents, there are strong reasons to believe discoveries made through animal studies may transfer to humans and thus provide profound insight into neurobiological processes supporting human spatial cognition. Furthermore, given the known role of the hippocampus in mnemonic processes, this field of research may not only shed light on the mechanisms underlying healthy human cognition but may also indicate processes impaired in amnesic diseases.

Edited by Ksenia Vinogradova