"Drive 200 yards, then turn right," says the car's computer voice. You relax in the driver's seat, follow the directions and reach your destination without error. It's certainly nice to have the Global Positioning System (GPS) to direct you to within a few yards of your goal. Yet if the satellite service's digital maps become even slightly outdated, you can become lost. then you have to rely on the ancient human skill of navigating in three-dimensional space.
Luckily, your biological finder has an important advantage over GPS: it does not go awry if only one part of the guidance system goes wrong, because it works in various ways. You can ask questions of people on the sidewalk. Or follow a street that looks familiar. Or rely on a navigational rubric: "If I keep the East River on my left, I will eventually cross 34th Street." The human positioning system is flexible and capable of learning. Anyone who knows the way from point A to point B--and from A to C--can probably figure out how to get from B to C, too.
But how does this complex cognitive system really work? Researchers are looking at several strategies people use to orient themselves in space: guidance, path integration and route following. We may use all three or combinations thereof. And as experts learn more about these navigational skills, they are making the case that our abilities may underlie our powers of memory and logical thinking.
Grand Central, Please
Imagine that you have arrived in a place you have never visited--New York City. You get off the train at Grand Central Terminal in midtown Manhattan. You have a few hours to explore before you must return for your ride home. You head uptown to see popular spots you have been told about: Rockefeller Center, Central Park, the Metropolitan Museum of Art. You meander in and out of shops along the way. Suddenly, it is time to get back to the station. But how?
If you ask passersby for help, most likely you will receive information in many different forms. A person who orients herself by a prominent landmark would gesture southward: "Look down there. See the tall, broad MetLife Building? Head for that--the station is right below it." Neurologists call this navigational approach "guidance," meaning that a landmark visible from a distance serves as the marker for one's destination.
Another city dweller might say: "What places do you remember passing? ... Okay. Go toward the end of Central Park, then walk down to St. Patrick's Cathedral. A few more blocks, and Grand Central will be off to your left." In this case, you are pointed toward the most recent place you recall, and you aim for it. Once there you head for the next notable place and so on, retracing your path. Your brain is adding together the individual legs of your trek into a cumulative progress report. Researchers call this strategy "path integration."
Many animals rely primarily on path integration to get around, including insects, spiders, crabs and rodents. The desert ants of the genus Cataglyphis employ this method to return from foraging as far as 100 yards away. They note the general direction they came from and retrace their steps, using the polarization of sunlight to orient themselves even under overcast skies. On their way back they are faithful to this inner homing vector. Even when a scientist picks up an ant and puts it in a totally different spot, the insect stubbornly proceeds in the originally determined direction until it has gone "back" all of the distance it wandered from its nest. Only then does the ant realize it has not succeeded, and it begins to walk in successively larger loops to find its way home.
Whether it is trying to get back to the anthill or the train station, any animal using path integration must keep track of its own movements so it knows, while returning, which segments it has already completed. As you move, your brain gathers data from your environment--sights, sounds, smells, lighting, muscle contractions, a sense of time passing--to determine which way your body has gone. The church spire, the sizzling sausages on that vendor's grill, the open courtyard, the train station--all represent snapshots of memorable junctures during your journey.
In addition to guidance and path integration, we use a third method for finding our way. An office worker you approach for help on a Manhattan street corner might say: "Walk straight down Fifth, turn left on 47th, turn right on Park, go through the walkway under the Helmsley Building, then cross the street to the MetLife Building into Grand Central." This strategy, called route following, uses landmarks such as buildings and street names, plus directions--straight, turn, go through--for reaching intermediate points. Route following is more precise than guidance or path integration, but if you forget the details and take a wrong turn, the only way to recover is to backtrack until you reach a familiar spot, because you do not know the general direction or have a reference landmark for your goal.
The route-following navigation strategy truly challenges the brain. We have to keep all the landmarks and intermediate directions in our head. It is the most detailed and therefore most reliable method, but it can be undone by routine memory lapses. With path integration, our cognitive memory is less burdened; it has to deal with only a few general instructions and the homing vector. Path integration works because it relies most fundamentally on our knowledge of our body's general direction of movement, and we always have access to these inputs. Nevertheless, people often choose to give route-following directions, in part because saying "Go straight that way!" just does not work in our complex, man-made surroundings.
Road Map or Metaphor?
On your next visit to Manhattan you will rely on your memory to get around. Most likely you will use guidance, path integration and route following in various combinations. But how exactly do these constructs deliver concrete directions? Do we humans have, as an image of the real world, a kind of road map in our heads--with symbols for cities, train stations and churches; thick lines for highways; narrow lines for local streets?
Neurobiologists and cognitive psychologists do call the portion of our memory that controls navigation a "cognitive map." The map metaphor is obviously seductive: maps are the easiest way to present geographic information for convenient visual inspection. In many cultures, maps were developed before writing, and today they are used in almost every society. It is even possible that maps derive from a universal way in which our spatial-memory networks are wired.
Yet the notion of a literal map in our heads may be misleading; a growing body of research implies that the cognitive map is mostly a metaphor. It may be more like a hierarchical structure of relationships. To get back to Grand Central, you first envision the large scale--that is, you visualize the general direction of the station. Within that system you then imagine the route to the last place you remember. After that, you observe your nearby surroundings to pick out a recognizable storefront or street corner that will send you toward that place. In this hierarchical, or nested, scheme, positions and distances are relative, in contrast with a road map, where the same information is shown in a geometrically precise scale.
Behavioral evidence also undermines the idea of a literal mental map. For one, map reading is not particularly easy. Children have to work at learning the skill, and many adults can live for decades in a city without being instantly able to find their residence on a map. Sketching a map of even a familiar town is a challenge for many people.
Perhaps people are more like the desert ant, which appears to memorize only what is necessary for its immediate trip, without creating anything like a complete map. We may deal with our daily routes from home to office and office to caf in a similar manner. The idea that humans and other animals rely primarily on a basic dead-reckoning approach to navigation attacks a widely shared prejudice among neurobiologists, who claim that mammals store spatial knowledge differently than lower animals. The conventional wisdom is that people create complex maps that include abstract entities and that are independent of the perspective of the person who is moving through a course--a kind of coherent overview that is in agreement with the coordinates of the real world. The ant knows only routes to and from its nest. It cannot take shortcuts from one foraging area to another--it must always go back to the nest first.
As they debate the extremes, researchers are homing in on a locational-memory model for humans that lies somewhere between "map in the head" and "learn by rote." Ranxiao Frances Wang of the University of Illinois and Elizabeth S. Spelke of Harvard University described such a model in 2002. Imagine, again, that you are on your first stroll through midtown Manhattan. As soon as you get off the train and as you wander, you take photographs of notable locations using a Polaroid camera. The first picture might show the hot dog vendor just up the block from the station, the second photograph a broad statue several blocks away, the next a striking cathedral one avenue over and so forth.
You number the snapshots as you advance, noting how you have gotten from one place to the next. If you walk down an unknown street and reach a location that seems familiar, you can review your collection of snapshots; if you have an image of the place, you can write down how you got there from the last location you photographed. All the while, as you journey onward, your brain is busy collecting images of unique locations and imprinting the paths that connect them, step by step, creating a denser and denser network.
When it is time to return to the station, you search your memory for pathways from image to image, piecing together a route back to the first picture. Like the ant, you are remembering only items that matter. Yet to save time, you may skip a snapshot and devise a more direct path from your current spot to the place shown two pictures earlier; unlike the ant, you are making creative and flexible use of your memory for locations. You remember many places, many routes, and you can formulate complex paths among them. And yet, in principle, this locational-memory model manipulates just two elements: places and routes. The model is powerful yet simple.
To assess whether human locational memory works in the way just described, the research group with which I worked at the Max Planck Institute for Biological Cybernetics in Tübingen, Germany, devised a test to track how people navigate through a virtual environment. The subjects sit in front of a color monitor that displays a computer-generated city called Hexatown, because its streets are laid out in hexagonal networks.
We asked test participants to observe a particular street in the town's network, which appeared to them as it would at eye level if they were standing in the middle of an actual street. We then asked them to "walk" around Hexatown and try to remember their routes. In addition to routine structures along the streets, two types of distinctive landmarks were available: unique local buildings placed at branches of streets and global references such as background mountains and tall buildings that were visible in the distance.
We then shifted the relations among the visual imagery by rotating the town and local landmarks while keeping the global landmarks fixed. We asked subjects, starting from a single point on a virtual street, to indicate which way the route they had previously learned now went.
Almost none of the subjects caught on to the rotation or used that as a clue; they continued to rely on their previous orientation strategies to try to rerun the original route. Some participants followed local landmarks and chose the same direction as before; the fact that the mountain range and tall buildings now rose in different places did not bother them. Without realizing it, they had rotated along with the town. Other people continued to orient themselves using the unchanged global landmarks, which obviously led them down completely new routes. They also failed to notice that the local landmarks--houses, squares, trees--were not the same as the ones they had seen before.
Does this mean that each of us relies on only one type of landmark and may not even have access to the others? To answer this question, we removed either the mountains or the unique buildings at crossroads. With little trouble, the participants switched to the other set of cues to rerun their course.
Apparently, our test subjects could orient themselves using local or global landmarks but decided to use only one type if both were available. It is not clear, however, why they did not notice that the two sets of landmarks had been rotated in relation to each other. To complicate matters, when we pointed out the shift after the tests were completed, most of them argued, often vehemently, that the relations among landmarks had not been altered.
We have more research to conduct. For now, we can conclude that human locational memories contain many individual bits of information, but we do not check to see if they are consistent with one another. As a result, contradictory bits may stand side by side, without confusing us. This observation indicates that cognitive maps are not similar to real road maps, because physical maps must be coherent.
The Subway of All Thought
What might cognitive maps look like in our heads, then? Perhaps they are like a graph, a collection of points and connections--something like a subway map. The points, or nodes, represent the unique landmarks we notice, and the lines between them correspond to actions that get us from one node to the next.
Note that on a good subway map, like that for Washington, D.C. [see illustration below], exact distances and accurately angled turns are unnecessary. The map only approximates the proportions of individual stretches and directions and puts nodes only in relative relation to one another. Exact scale and geographic rigor would add extraneous details that needlessly confuse navigational needs. They also eat up lots of memory. And new lines can be added without having to adjust all the details of the entire map.
Using a mental graph akin to a subway map, we can easily advance along a chain of landmarks to navigate from start to finish. In Manhattan, when it is time to return to the train station, we can retrace a step-by-step route or devise a new way to move directly from point C to point A. We can use guidance, path integration or route following (or some combination) to reach our destination, and we do not have to burden our brains with details that do not help us advance on our course.
Humans have a multitude of cognitive maps in their heads. Our locational memory has not changed for millions of years--it has simply refined the original principle. Indeed, philosophers, scholars and brain researchers have long suspected that spatial orientation is more than a special skill--it may be one evolutionary root of memory or thought itself. For example, Cordula Nitsch and her colleagues at the University of Basel in Switzerland showed in experiments with gerbils that increasing levels of damage to the hippocampus, a deep and ancient brain structure, increasingly impaired both the animals' spatial orientation and memory retention in navigating a course they had previously mastered.
The fact that we typically memorize locations better than abstract items of information is not just a sign of the key role of locational memory for our general ability to make a mental note of objects in our surroundings. In the 18th century philosopher Immanuel Kant had already listed the ideas of space, time and causality as the fundamental building blocks of human intelligence that did not stem from experience. According to Kant, humans simply cannot not think spatially. In the mid-1900s Nobel Prizewinning behavioral researcher Konrad Lorenz proposed that the complex three-dimensional environments of the first arboreal primates provided a strong impetus for the development of higher cognitive skills. And we see today that many of the idioms we use in daily speech have spatial roots: we "get oriented" to new situations, try to "find ways out" of our problems, and ask colleagues to "walk us through" proposed plans.
If spatial references readily transfer to nonspatial information, then the graph model can transfer to nongeographic tasks as well. To make a cake, you have to carry out a series of actions. You measure the ingredients, mix them together, fill the cake pan. Each step is a node, and the work you must do to get from step one to step two is the line connecting them. This baking graph is flexible and expandable. Some recipes call for eggs, which requires an additional step between "measure" and "mix"--specifically, cracking the eggs. You may have learned this skill in another context--making an omelet--but you add it to the cake-baking repertoire. In a similar way, a first-time visitor to Manhattan adds segments to his or her graph of how to get around from information gleaned from other contexts--the sun rises in the east, which indicates north, and a shop owner notes that Central Park is north, up Fifth Avenue, from Grand Central Terminal.
It is not inconceivable that over the course of human evolution a memory structure developed for spatial orientation--one that was later employed for other cognitive functions. The uses to which lower animals apply spatial cognition implies as much. Or to put it more provocatively: in the animal kingdom, spatial cognition is the most widespread form of thought.