Humans, just like all other animals, face the same problem every day: how do we get around the world? I don't mean how do we walk, swim, crawl, or fly. I mean, how do we navigate? If I leave in search of food, how do I find my way back home?
Here's one method I might use to find my way back home. First, I'll use the odometer in my car to figure out how many miles I've driven. I'm not going to be driving straight, though, I'll be taking turns - so I'll have to use some basic trigonometry to calculate the total distance to my house at each mile. For example, if I drive three miles , turn ninety degrees, and then drive four miles, then I'll know that home is five miles away - even if I don't know what direction I've driven, I can draw a five mile radius around my position, and I'll know that my house is somewhere along that circle. That takes care of distance. To figure out the direction I need to drive to get back home, for each mile I drive on my outbound trip, I can figure out the location of the sun in the sky relative to home's location and my current location, and, again, use some basic trigonometry to figure out what direction I need to drive to get back home. In addition, I'll have to use some sort of internal chronometer - a mental clock - so that I can correct for the changing location of the sun over time.
Indeed, this is one mechanism for navigation that humans use. And so do gerbils, and geese, chickens, and even desert ants. We don't actually sit there with a pad of graph paper and a protractor and recalculate our distance and position in any explicit way - rather, it all takes place below the level of conscious awareness. If an individual knows two of the angles and the distances between three objects, he or she can calculate the third angle and distance, using principles of basic Euclidean coordinate geometry. Some people call this process "dead reckoning," and in the scientific literature, it's referred to as "path integration."
There's a problem, though: path integration is subject to cumulative error. Over the course of one hour, for example, the estimate of direction can be off by as much as 10%, and the direction estimate can be off by as much as two degrees. Also, path integration requires continuous memory of one's position relative to home: if an individual becomes disoriented, or forgets, the ability to path integrate becomes useless. To reorient themselves, to regain their bearings, animals (human animals included) need to utilize enduring representations of the environment. For example, animals can rely on their viewpoint-dependent memory of visual scenes. In particular, we can use visual landmarks - items or features of the visual scene that tend not to change over time, such as signs, buildings, or even certain trees. It's one thing to assess whether or not an animal can find its way home. It's another problem entirely to determine the cognitive mechanism in place that allows the animal to get home in the first place. That, in a sentence, is the problem that cognitive psychologists must grapple with, every day.
Ask most people how they manage to navigate the world, and many will tell you that they think about the space around them in map-like terms. Perhaps they will explain that they imagine a bird's-eye view of their neighborhood. This, they will insist, allows them to get to the store from the office, even if they have never driven that particular route before (they usually drive to the store from home, never from the office).
For most of the past century, cognitive psychologists would have argued that human intutions happened to be correct when it came to the existence of a "map within the mind." Our story starts in the 1940s, with a group of rats at UC Berkeley in the laboratory of experimental psychologist Edward Tolman...
Laboratory rats can quickly learn their way around mazes, and once they do, they can move between individual locations within the maze quickly and efficiently. In the typical maze, a hungry rat is placed at the entrance and wanders around until he or she eventually finds the hidden food, and eats. As this process is repeated day after day, the rat makes fewer and fewer errors and the time it takes to reach the food gets shorter and shorter - presumably because he or she learns the proper routes.
One particularly useful maze that researchers use to test navigational abilities in mice and rats is the Morris water maze. The animal is put into the round tank and it begins to swim. In one spot, there is a submerged platform which will allow the animal to comfortably stand without needing to swim or tread water to keep afloat. This would be pretty easy in clear water, so the tank is filled with an opaque milky liquid. The walls of the tank have various geometric markings on them in different places. The question is: after being introduced to the maze and finding the location of the platform during training, will the animal be able to use the navigational cues given by the different geometric markers to swim directly towards the platform during testing? Importantly, the tank is circular, so corners and edges can not be used to aid in navigation - the only cues available are the geometric markings.
Indeed, rats reliably learn to swim directly for the platform, as long as the geometric markings remain constant. In 1948, Tolman used experiments like this and others to infer that rats and people constructed cognitive maps in their minds. He wrote, "We believe that in the course of learning something like a field map of the environment gets established in the rat's brain... And it is this tentative map, indicating routes and paths and environmental relationships, which finally determines what responses, if any, the animal will finally release." The idea of the cognitive map became well-entrenched in cognitive psychology. It certainly is intuitive.
Neurophysiological experiments were designed that allowed researchers to monitor the activity of individual neurons in rats while they navigated a maze. Neurons in the hippocampus activate reliably in these sorts of tasks, with each neuron firing in response to a specific place in the environment (these have been called "place cells"). Supporting this theory is the finding that damage to the hippocampus impairs performance on the water maze. By the late 1980s, cognitive neuroscientists came to believe that the cognitive map was stored within the hippocampus.
For more than thirty years, the existence of the cognitive map was generally accepted in psychology. But by the early nineties, the cognitive map began to fray.
It was certainly possible that the types of mental representations that rats used while navigating were like real maps, as Tolman thought, containing information about the locations of places and the relationship of each place to each other place, independently of the individual's viewpoint. In this sort of model, the cognitive map consists of a "bird's-eye view" of the environment.
There is another possibility, however. It is possible that the mental representations activated by these sorts of navigation tasks are like photographs or snapshots, which simply capture the appearance of each place from a particular point-of-view. Do rats navigate by using a cognitive map or by using a series of viewpoint-specific visual "snapshots"?
To determine which type of mental representation they used, in 1991, researchers constructed a new kind of water maze. It was already known that given free access to all vantage points from within the maze, rats could navigate to the platform from anywhere else in the maze. What if the researchers restricted the rats' access to one portion of the Morris water maze during the training phase? This would prevent the rat from forming any viewpoint-specific snapshots from anywhere within the blocked-off space. Then, during testing, they would remove the barrier (note the light blue barrier in the top left corner of the diagram) and drop the rat into the maze in the space where the barrier used to be.
If the cognitive map hypothesis was correct, then the rats should have been able to successfully navigate directly to the platform from the part of the maze in which they had not been able to form any visual "snapshots." However, if rats rely on viewpoint-specific "snapshots" of the visual scene in order to navigate, then when dropped into the new area, they ought to behave as if they're in an entirely new maze. If cognitive maps didn't exist, then the rats should have taken significantly longer to find the platform. Indeed, rats placed into the unfamiliar part of the maze were far less successful than when they were placed into the familiar parts of the maze.
It seems unlikely that rats would be the only animals to toss aside the cognitive map. Another set of critters that have been studied in depth for navigation are insects - especially ants and bees. When bees are given a task in which they must find some hidden food, do they form a cognitive map or take viewpoint-specific snapshots? In a 1997 study, researchers separated a bunch of bees into two groups, and each had different forms of training in an arena full of objects. The first group was able to fly around the entire arena until it found the hidden food, as they would in a natural foraging environment. Then, during test, no matter what side of the arena the bees were released from, they were able to immediately locate the food again. Just like the rats. But a second group of bees were only trained to approach the food from one specific location. When released from the same spot during testing, those bees were successful, as would be expected. The critical question is what would the bees do who had been trained on one route, but were released from a new location? If bees rely on viewpoint-specific snapshots to navigate, then those bees should have taken much longer to find the food, compared with those tested on the same route that they had trained. That is, of course, exactly what happened.
Both bees and rats learn to navigate a new environment not by forming cognitive maps, but by using viewpoint-dependent snapshots of scenes within the environment. And not just bee and rats; subsequent studies have verified that lots of other non-human animal species also have a view-dependent scene recognition mechanism that they use to navigate, rather than a map-like representation. The next step was to look at humans: do we use cognitive maps, or do we use viewpoint-dependent snapshots?
In 1998, a pair of cognitive psychologists named Dan Simons and Ranxiao Wang came up with a clever way to look at this in human adults. Experiment participants sat down on a table, and on the table was a set of objects: a brush, a mug, goggles, a stapler, and scissors. After viewing the scene, a curtain would lower over the table. Behind the curtain, one of the objects was moved. Then the curtain was raised, and the participants had to identify which of the five items was moved. The trick was that for some of the trials, the entire array of objects would be rotated as well, though each object would be in the same place relative to each other object (except for the one that had been moved). If humans use cognitive maps to remember visual scenes, then the rotation of the table should not negatively affect performance: the participants should have been as good at noticing which item was moved when the table was rotated as when the table was left alone. When the table was not rotated, as expected, it was easy for the human participants to indicate which of the five objects had been moved. However, their performance suffered when the table was rotated. In other words, when looking at the table from a different viewpoint, they were unable to determine which item was in a different spot relative to each other item. Human adults, it appeared, use a view-dependent scene recognition system as well! They take visual snapshots, rather than forming a cognitive map. By changing the viewpoint, the human participants might as well have been looking at an entirely new array of objects. Similar experiments have been conducted with children and even with young infants. Together, these studies show that this system comes online quite early in life, and persists relatively unchanged through development.
If humans indeed have the same cognitive mechanism for representing scenes as bees and rats and other animals, then it is reasonable to ask if this specialized mechanism also has a specialized location in the brain. Studies of rats indicated that representations of scenes are localized in the hippocampus and the surrounding area. If humans form the same sorts of "snapshots," then similar areas ought to be activated by similar tasks in humans as in rats. To test this prediction, cognitive neuroscientists Epstein and Kanwisher conducted an fMRI experiment in which human adults simply viewed pictures of scenes, faces, and objects. What they found was that one particular location in the brain - indeed, within the hippocampus! - responded to scenes, very weakly in response to objects, and not at all to faces. They named it the parahippocampal place area, which has become known as the PPA (indicated in each of the nine participants below by the yellow arrows).
Epstein, Kanwisher, and others did more fMRI experiments to further clarify the rule of the PPA. They found that it reliably activated in response to scenes, whether indoor or outdoor, and whether those scenes are empty or full of objects. They do not highly activate, on the other hand, for things that are typically located within scenes (such as furniture within a room) in the absence of surrounding surfaces (such as the walls of a room). Taken together, the evidence is overwhelming. The PPA does seem to be the homologous region in humans, for the "place neurons" In the rat hippocampus.
And so it was that the end of the millennium brought with it the end of the cognitive map. This is not to say, of course, that humans can't use maps. The critical difference is that our internal representations of the environment are not map-like. Luckily, my GPS doesn't force me to use a snapshot matching system to get around Los Angeles!
Tolman, E. (1948). Cognitive maps in rats and men. Psychological Review, 55 (4), 189-208 DOI: 10.1037/h0061626
O'Keefe J, & Speakman A (1987). Single unit activity in the rat hippocampus during a spatial memory task. Experimental brain research. Experimentelle Hirnforschung. Experimentation cerebrale, 68 (1), 1-27 PMID: 3691688
Whishaw IQ (1991). Latent learning in a swimming pool place task by rats: evidence for the use of associative and not cognitive mapping processes. The Quarterly journal of experimental psychology. B, Comparative and physiological psychology, 43 (1), 83-103 PMID: 2017576
Collett, T., & Rees, J. (1997). View-based navigation in Hymenoptera: multiple strategies of landmark guidance in the approach to a feeder Journal of Comparative Physiology A: Sensory, Neural, and Behavioral Physiology, 181 (1), 47-58 DOI: 10.1007/s003590050092
Simons, D., & Wang, R. (1998). Perceiving Real-World Viewpoint Changes Psychological Science, 9 (4), 315-320 DOI: 10.1111/1467-9280.00062
Epstein R, & Kanwisher N (1998). A cortical representation of the local visual environment. Nature, 392 (6676), 598-601 PMID: 9560155
Images: Downtown LA copyright the author. Morris water mazes with permission of Justin Wood. Bee experiment schematics are the brilliant artwork of the author. fMRI images from Epstein and Kanwisher (1998).