A tiny, pixilated soldier dodges past burning embers and ruined walls. His guide, a young boy watching through a computer monitor, knows that just ahead, beyond a darkened doorway and a hairpin left turn, the soldier will find a floating white medical kit to nourish and soothe his battered body. He will recharge, then navigate his way through an extensive labyrinth of corridors to the next level of the maze.
The boy playing the video game nudges his joystick, guiding the soldier efficiently through countless rooms. He knows this virtual world well, and has an intimate understanding of its topography. In his mind’s eye the bitmapped patterns and flashing lights become three-dimensional hallways, staircases, and doors.
In the early years of the 20th century, behaviorists predicted that spatial memory, our knowledge of physical places, depended on associating landmarks with rich perceptual input. Most of us are able to navigate the corridors of a familiar house in the dark, recognizing features by touch and texture, by smell and sound. But when that boy wiggles his joystick, sending the soldier into a secret doorway for extra ammunition, he is navigating a virtual world, without the type of spatial memory formed by intimate sensory experiences. Scientists have learned that spatial memory is robust, and that the brain structures responsible for this sense of the world do more than simply build maps from sensory information. New evidence suggests that our sense of space is an integral part of who we are, and how we relate to the world.
Map Making and the Brain
Humans have an affinity for maps and map-making, so we have a good understanding of how maps can be used to represent spaces. Most of us can draw a crude sketch of the house that we lived in as a child, or of the streets that we take to work. Early theories about spatial memory drew upon this cartographic sense to suggest that our ability to navigate our environment resulted from the construction of a “cognitive map,” an internal representation of the real world. Theories varied on the nature of this internal representation. Some suggested that the cognitive map could be egocentric, or centered on the self – objects in the map would be located relative to the head, and the map representation would change as the self moved through the landscape. Other theories proposed an allocentric map, one based on an external frame of reference. In an allocentric view, objects have a fixed location on the map independent of the observer, and self-motion does not affect the map. In this view, the self appears something like a Monopoly token, marking the position and direction of movement of the body relative to its environment.
Both egocentric and allocentric views of the cognitive map offered a prospect that tantalized brain researchers from the beginning: they were testable theories about how the brain represented the external world. Scientists recognized that any kind of cognitive map would have certain characteristics, telltale signs that would show up in the neurons that created the map. An allocentric map would be strongly correlated with the physical environment; if you walked into the same room twice, you would expect that your cognitive map of the room would be largely the same both times. If the cognitive map is egocentric, it should change as the position of the self in the environment changes. The testable difference is that an allocentric map would be stable, but an egocentric map would be constantly changing. While scientists debated the finer points of the cognitive mapping hypothesis (were the maps based on Cartesian or polar coordinates? Were they measured in units of footsteps, or the time it took to traverse spaces?) the race was on to find a structure in the brain that bore the tell-tale marks of a mapmaker.
Finding the Neural Cartographers
The first neurobiological evidence for a cognitive map was presented by John O’Keefe and John Dostrovsky at University College London. Like most scientists researching spatial memory today, O’Keefe and Dostrovsky were struggling to explain how rats rapidly learned to navigate mazes. The team studied a seahorse-shaped structure called the hippocampus that curled towards the center of the brain (Figure 1). Rats with hippocampal damage were very slow at finding their way through mazes, so the structure was clearly important for spatial navigation. But was the hippocampus the long sought-after mapmaker?
In 1971, O’Keefe and Dostrovsky reported that they had observed neurons in the rat hippocampus that were very active in certain well-defined regions of a maze. They called these neurons “place cells,” because their activity was correlated with places, and not with smells, sounds, or other features of the environment. The parts of the maze that activated any particular place cell were called that cell’s “place field,” in analogy to receptive fields in the sensory systems. Place fields seemed to cluster together in the environment – that is, parts of the maze that were close to each other would activate place cells in the hippocampus that were close to each other. All the evidence seemed to indicate that place cells were the neural incarnation of the cognitive map (Figure 2).
Scientists plunged into research on place cells enthusiastically. The decades following O’Keefe and Dostrovsky’s report witnessed a bounty of work describing the length of time that a place cell could remember it’s place field, the environmental cues that were necessary for establishing place fields, and the biochemical properties of place cells that were required for place field formation. Despite this flurry of activity, some difficulties with the research persisted. While place fields clustered in small groups, they didn’t seem to be organized in any larger pattern. And while a place cell might be active when the animal was standing in its place field, that cell would be silent in any other part of the environment. Shouldn’t a cognitive mapmaker have some understanding of topography, of how small parts of the world relate to other parts of the world? And shouldn’t a cognitive map show you how to get from one place to another, instead of just presenting a big cellular “you are here” sign? Many theories emerged to bring the wily characteristics of the hippocampus in line with its responsibilities as the brain’s cartographer, but by the early 1990s it was clear that a new understanding of the hippocampus’ role in spatial memory was needed.
New Views On Space
One of the biggest challenges to the cognitive map theory of hippocampal function is that while rats with hippocampal damage just get lost a lot, humans with hippocampal damage exhibit complex, non-spatial forms of amnesia. Amnesiac patients can often recall events in the distant past more easily than the recent past; they may forget your name shortly after they meet you, but they can vividly recall the drawings in their kindergarten classroom. Brain imaging studies have shown conclusively that humans, like their furry rodent counterparts, use their hippocampus when mentally navigating city streets, or when exploring video game mazes, but so far there has been no way to determine whether humans have hippocampal place cells. A new theory of hippocampal function must reconcile the evidence from rats, which suggests that the hippocampus is principally concerned with spatial navigation, with the evidence from humans, which suggests that the hippocampus is generally associated with long-term memory formation.
Howard Eichenbaum, a scientist at Boston University, has proposed a theory that might explain both bodies of evidence. Eichenbaum conceives of the hippocampus as a “memory space” that encodes many different aspects of daily life. In particular, he argues that the hippocampus records the regularities of an experience, those features that are dependable and useful in making sense of the world. For example, when you recall the smell of fresh baked cookies in your grandmother’s kitchen, you are associating a scent with a place. Both are useful to remember together, because as a pair these memories form a meaningful description of what happened in a particular spot and a particular context.
Eichenbaum calls associations that are meaningful in a behavioral context “events,” and has suggested that the main role of the hippocampus is to collect these events and link them into behavioral episodes. Episodes are threads of events that are linked because one event follows another in time. After tasting your grandma’s cookies, you might go into the family room to play a game with your little brother. Drowsy and content, you could then make your way to your bedroom, snuggling into a warm bed for a snooze. The memory space theory suggests that your hippocampus pays attention to all of the events that happen, links them together into short episodes, and then begins to make connections where the episodes overlap.
All of this linking and associating is a good way to build an understanding of spatial relationships. The path from the kitchen through the family room to your bedroom would be easy to remember after you’d experienced events in these places. Remembering the place in the context of behavioral episodes would also let you remember how each place relates to your life, and of how the events of your life relate to each other. That makes the hippocampus much more than a neural mapmaker; in the Eichenbaum model, the hippocampus is a cognitive historian that brings the events of life together into meaningful relationships.
Researchers universally agree that whether the hippocampus is a mapmaker or an historian, it’s only one player in the process of memory formation. The fact that amnesiac patients with hippocampal damage can recall childhood memories suggests that long-term memories themselves are stored elsewhere in the brain. That observation suggests that the hippocampus acts to package memories into a form that can be readily stored, to clean up the details and retain the important features of an event — a process called consolidation. The Eichenbaum hypothesis is a model for how consolidation might work, emphasizing episodic sequence as the thread that binds memories together.
If the hippocampus is not a cognitive mapmaker, then we’re still left with a quandary about where in the brain to find such a mapmaker. When a taxi driver plans a route through crowded city streets, when a child guides a character through a video game, when we move through familiar places in our daily lives, we can and do visualize the layout of spaces in a mental map. While Eichenbaum’s theory suggests that we don’t actually use spatial visualization as a principal way to help us get around, understanding how we generate abstract representations of real and virtual spaces is a fundamental part of understanding how thought and abstraction work in general. In the coming years, new research may cast the hippocampus in more and varied roles in the marvelous neural interplay that allows us to navigate our world.
Eichenbaum, Howard, Dudchenko P., Wood E., Shapiro M., and Tanila, H.: The Hippocampus, Memory, and Place Cells: Is It Spatial Memory or a Memory Space? Neuron 23: 209-226, 1999.