Navigation In Spatial Environments Research Paper

2. The Hippocampal Formation Has Been Implicated As A Possible Location For The Cognitive Map

Studies of individuals who have sustained focal brain damage due to illness or injury have revealed certain brain regions that may contain the postulated map-like representations. For example, experimental studies in rats have suggested that a brain region known as the hippocampal formation may be critical, since damage to this area causes deficits on spatial learning tasks (O’Keefe and Nadel 1978). In addition, humans with damage to this region sometimes show selective deficits in topographical orientation (Habib and Sirigu 1987, Maguire et al. 1996a). These individuals are unable to learn their way around in a new environment, and when tested formally, they are impaired in making correct judgments about the relative distance or direction between different locations (Maguire et al. 1996a). This topographical disorientation can sometimes appear in the absence of other detectable mnemonic or perceptual deficits. That is, the people can recognize and name landmarks, and can even learn and remember the identity of new landmarks, but they cannot learn effectively about the spatial relations between them, or how to navigate from one to the other.

Complimentary evidence for the importance of the hippocampal region has come from brain imaging studies that employ navigation tasks (Aguirre et al. 1996, Maguire et al. 1996b, 1997, 1998). In one study (Maguire et al. 1997), experienced London taxi drivers were asked to plan routes from one location to another within the city. For comparison, the drivers were also asked to perform several control tasks, such as describing famous landmarks, or recounting the plot sequence from well-known films. Relative to the control tasks, the route-planning task caused relatively selective activation of the right hippocampus.

Despite the apparent importance of the hippocampal formation for learning about new environments, recent evidence suggests that navigational memory for environments learned long ago may reside in cortical areas other than the hippocampus. For example, a study by Teng and Squire (1999) reported on an individual who had sustained massive bilateral damage to the temporal lobes, due to an episode of encephalitis. As is well documented, large temporal lobe lesions like this cause a global amnesia for episodes, events, and places which are experienced after the damage, as well as for those which occurred for some time prior to the damage. Of relevance here, the patient reported by Teng and Squire (1999) had normal recall for the urban area in which he had grown up, but had moved away from, nearly 40 years earlier. He was able to accurately describe how to travel from one location to another within this environment, and when asked to imagine himself positioned in a certain location and directional orientation, he could accurately point in the direction of various other landmarks. Thus, these data indicate that the ability to recall and utilize topographic information for an environment learned long ago must reside in some area other than (or in addition to) the hippocampus.

3. The Retrosplenial (Posterior Cingulate) Cortex Has Also Been Implicated As A Possible Location For The Cognitive Map

If topographical information for an environment learned long ago eventually becomes independent of the hippocampal formation, this raises the question as to what other brain region may take over these memories. Recent evidence has suggested that the retrosplenial cortex may be one important region. Takahashi et al. (1997) have reported on three patients who showed selective topographic amnesia after sustaining damage to the right retrosplenial region. Importantly, the disorientation experienced by these people was equally severe for both familiar as well as new environments.

Additional evidence for the importance of the retrosplenial cortex comes from the brain imaging study outlined above (Maguire et al. 1997); retrosplenial cortex was another brain area which was selectively activated during the taxi drivers’ route planning.

4. Investigations Of Single Cell Activity In The Hippocampal Formation And Retrosplenial Cortex Have Revealed Clues About The Cellular Basis Of The Postulated Cognitive Map

In order to gain insight into the detailed nature of the postulated cognitive map, many investigators, beginning with O’Keefe and colleagues (e.g., O’Keefe and Dostrovsky 1971, O’Keefe and Conway 1978) have looked at the activity of individual neurons in the relevant brain regions of experimental animals, to see what kinds of spatial signals may exist. These investigations have revealed two main types of spatial cell.

4.1 Place Cells

One type, known as place cells, fires whenever the rat is in a particular location. For example, one cell may fire at a high rate whenever the rat is in the northeast corner of a square enclosure, and be completely inactive when the rat is in any other location. Each of these place cells has its own region of high firing so that, as a whole, the population of hippocampal place cells provides an ongoing representation of the animal’s current location; each location the rat occupies is coded by activity in a unique subset of the hippocampal cells. Cells of this general type have been found throughout the hippocampal formation (O’Keefe and Dostrovsky 1971, Sharp and Green 1994, Taube 1995) and also in the retrosplenial cortex (Cho and Sharp 2001).

It is not completely clear how these location-specific signals are generated. One possibility could be that they simply respond to the presence of landmarks (environmental stimuli) which are available in the preferred region. This possible explanation is sup-ported by early data which showed that the location of the hippocampal place cell firing fields can be controlled by the location of environmental cues. For example, in one study (Muller and Kubie 1987) cells were recorded while rats navigated in a gray cylindrical apparatus. The wall of the cylinder was equipped with a single, white cue card that acted as a landmark. When the card was rotated during separate probe sessions, the location of each of the place cell fields rotated as well, by the same amount.

Additional data, however, showed that the place cell signals are more complicated than suggested by this simple sensory explanation. For example, in the study outlined above (Muller and Kubie 1987) it was found that when the white card was removed entirely, the place cells could still retain location-specific firing patterns. In this case, each place cell showed a firing field that was the same size, shape, and distance from the wall as it was in the presence of the white card. However, the angular location of the fields was unpredictable in this case. During any one recording session, each cell fired in just one location, but the location that it chose could not be predicted in advance.

Data of this sort suggest that the cells utilize a process known as path integration (ded reckoning), at least in part, to generate the place cell signals (e.g., McNaughton et al. 1995). Path integration refers to the use of one’s own movement to track position. The behavior of the place cells in the no-card probe sessions could be explained using this idea by suggesting that, first, when the animal is introduced into the chamber at the start of the session, the place cell firing pattern is randomly set to represent some particular location. Thus, since there is no polarizing cue, when the rat is initially placed down next to the cylinder wall, place cell activity appropriate to, for example, the northwest region of the inner wall may be randomly ‘cued’ up. Then, after this initialization process, when the animal begins to move, the system somehow uses information about the speed and direction of movement to continuously update the representation of the animal’s current position.

The fact that the place cells are controlled by both environmental landmarks and path integration is remarkably compatible with introspective evidence about our own navigation. Thus, in a familiar environment, if one becomes momentarily confused about one’s location, it is usually possible to simply look around for a familiar landmark, to become reoriented. However, if familiar landmarks become unavailable (such as might happen after dark, or when one has ventured into new territory), then it is still possible to have some sense of where one is, just based on information about where the journey began, and the subsequent direction and distance traveled.

4.2 Head Direction Cells

The other type of spatial cell is known as the head direction cell (Ranck 1984, Taube et al. 1990a). Each of these cells fires whenever the rat faces one particular direction. For example, one cell may fire whenever the rat faces southeast, while another fires only when the rat faces north, etc. Thus, these cells are the compliment of the place cells, in that they provide an ongoing representation of the current directional heading, rather than location. These cells were originally discovered in the subicular complex of the hippocampal formation (Ranck 1984), and have since been discovered in several other areas as well, including the retrosplenial cortex (Chen et al. 1994, Cho and Sharp 2001). As is the case for the place cells, the head direction cells have also been found to utilize both visual landmarks as well as path integration to keep track of the animal’s directional heading (Taube et al. 1990b).

5. Neural Network Models Ha E Been De Eloped To Explain The Firing Properties Of The Place Cells And Head Direction Cells

Several neural network models have been developed to attempt to replicate the spatial firing characteristics of place cells and head direction cells.

5.1 Models Of Place Cells

Many of the place cell models have focused on the fact that these cells are influenced by environmental cues, and they attempt to explain the properties of place cells as due to convergent sensory inputs (e.g., O’Keefe and Burgess 1996, Sharp 1991, Treves et al. 1992, Zipser 1985).

One obvious limitation of these models is that they do not account for the path integration properties shown by these cells. For this, McNaughton and colleagues have developed a model in which movement information (along with environmental cues) serves as an input to the hippocampal system (McNaughton et al. 1995, Samsonovich and McNaughton 1997). Specifically, the place cells themselves project onto another postulated cell layer which also receives input about the current direction and speed of movement. Thus, each cell in this layer fires to a particular combination of place, movement, and direction. These cells then project back onto the place cells themselves, and are connected so that they induce firing in the appropriate ‘next’ set of place cells (i.e., those which represent the location which will result from the current location, direction, and speed of movement).

5.2 Models Of Head Direction Cells

Several models have been developed to simulate the characteristics of head direction cells (Redish et al. 1996, Sharp et al. 1996, Skaggs et al. 1995, Zhang 1996). These models all have the same basic properties. They all begin with the idea that the head direction cells themselves are linked together to form a stable attractor. In this network, cells that are meant to represent different directions (e.g., north vs. south) have strong, mutual, inhibitory connections, while cells meant to represent similar directions are linked through excitatory connections. This connectivity ensures that, at any one time, the system will settle into a pattern of activity in which cells which represent one directional range will be active, while all other cells are silent.

To implement the path integration abilities, the system also receives information about angular head movement (such as might be provided by the vestibular system, or motor command signals). Cells carrying this information project onto another postulated layer which also receives input from the head direction cells themselves. Each of these cells, thus, fires for a unique combination of current directional heading and angular movement. This combination allows for the prediction of the next directional heading (e.g., a current heading of north, along with a 90-degree clockwise turn will result in a directional heading of east). These cells then project back onto the head direction cells in such a way that they drive the next appropriate set of head direction cells.

The head direction cells are also assumed to receive input from sensory cells, so that they can be ‘set’ by environmental cues, as described above.

6. There Is Still Much To Be Learned About The Cognitive Map

The discovery of place cells and head direction cells constitutes a major step toward an understanding of the neural computational basis of navigation. However, there are critical aspects of the story which are yet unknown. In particular, it is not known how these place and direction signals are used to generate actual goal-directed behavior.

Although there has been remarkably little empirical work on this problem, several neural modeling approaches have been used to attempt to provide a conceptual framework for this aspect of navigation.

In one type of model it is assumed that the information necessary for route planning resides in the connectivity of the place cells themselves (Blum and Abbott 1995, Hetherington and Shapiro 1993, Muller et al. 1996). It is assumed that as an animal explores an environment, place cells which fire successively along particular paths taken by the animal will become strongly connected, due to a Hebb-like synaptic plasticity. In the Muller et al. (1996) model, this connectivity is used to find the shortest path between the place cell representations of the animal’s current position and a desired goal. In the Blum and Abbott (1995) model, the strong connection between two cells along a pathway is assumed to cause the firing of place cells at any one moment to ‘jump ahead’ a bit, so that cells corresponding to the ‘next’ location are active. Moreover, this connectivity between place cells is presumed to be enhanced by previously experienced reward. During recall, the rat selects an optimal trajectory by traveling through the path represented by the activity of the currently active place cells. Due to the experience on previously rewarded trials, these place cells represent the next step in the optimal path toward reward.

One limitation of these models is that they do not specify how the activity of the place cells themselves can be translated into actual behavior. Additional neural machinery would be necessary to calculate how to move from the current position to the one indicated by the trajectory-related place cell activity.

A model which attempts to provide the necessary route-generating machinery has been provided by Brown and Sharp (1995). In this model, both place cells and head direction cells project onto a layer of motor cells which drive behavior. Initially, the strengths of these connections are set randomly. When the rat encounters reward, however, recently active connections between place head direction cells and motor command cells are strengthened. This means that behaviors that lead to the goal from a particular position are strengthened.

One limitation of this model is that it does not work over long distances. This is true since the only connections that can be strengthened are those that were active a short time before the animal reached the goal.

A postulated solution to this problem has been provided by Foster et al. (2000). In this model, a temporal distance learning module has been added. Over the course of the animal’s exploration this model eventually learns the reward value of each step the animal takes, regardless of where it is in the environment. The output of this module is then used to guide the actual motor learning.

7. Summary

In summary, the remarkable navigational abilities demonstrated by both animals and humans are thought to depend on a sophisticated, internal representation of the environment, sometimes referred to as a cognitive map. Two brain regions that have been implicated as possibly containing the cognitive map are the hippocampal formation and the retrosplenial cortex. It appears that these two regions somehow work together to help orchestrate navigational abilities. Specifically, the hippocampus seems to play a necessary role during initial learning about a new environment, while the retrosplenial cortex may be necessary at all stages. The discovery of place cells and head direction cells in both these regions has provided some clues about the details of the cognitive map, but it is still not clear how these cells actually help to guide navigational behavior.

Bibliography:

1. Aguirre G K, Detre J A, Alsop D C, D’Esposito M 1996 The parahippocampus subserves topographical learning in man. Cerebral Cortex 6: 823–29
2. Blair H T, Sharp P E 1996 Visual and vestibular influences on head direction cells in the anterior thalamus of the rat. Behavioral Neuroscience 110: 1–18
3. Blum K I, Abbott L F 1995 A model of spatial map formation in the hippocampus of the rat. Neural Computation 8: 85–93
4. Brown M A, Sharp P E 1995 Simulation of spatial learning in the Morris water maze by a neural network model of the hippocampal formation and nucleus accumbens. Hippocampus 5: 189–97
5. Chen L L, Lin L H, Green E J, Barnes C A, McNaughton B L 1994 Head direction cells in the rat posterior cortex. I. Anatomical distribution and behavioral modulation. Experimental Brain Research 101: 8–23
6. Cho J, Sharp P E 2001 Head direction, place, and movement correlates for cells in the rat retrosplenial cortex. Behavioral Neuroscience 115: 3–25
7. Foster D J, Morris R G M, Dayan P 2000 A model of hippocampally dependent navigation, using the temporal difference learning rule. Hippocampus 10: 1–16
8. Gallistel C R 1990 The Organization of Learning. MIT, Cambridge, MA
9. Habib M, Sirigu A 1987 Pure topographical disorientation: A definition and anatomical basis. Cortex 23: 73–85
10. Hetherington P A, Shapiro M L 1993 A simple network model simulates hippocampal place fields: II. Computing goaldirected trajectories and memory fields. Behavioral Neuroscience 107: 434–43
11. Maguire E A, Burgess N, Donnett J G, Frackowiak R S J, Frith C D, O’Keefe J 1998 Knowing where and getting there: A human navigational network. Science 280: 921–24
12. Maguire E A, Burke T, Phillips J, Staunton H 1996a Topographical disorientation following unilateral temporal lobe lesions in humans. Neuropsychologia 34: 993–1001
13. Maguire E A, Frackowiak R S J, Frith C D 1996b Learning to find your way: A role for the human hippocampal formation. Proceedings of the Royal Society of London B 263: 1745–50
14. Maguire E A, Frackowiak R S J, Frith C D 1997 Recalling routes around London: Activation of the right hippocampus in taxi drivers. Journal of Neuroscience 17: 7103–10
15. McNaughton B L, Barnes C A, Gerrard J L, Gothard K, Jung M W, Knierim J J, Kudrimoti H, Quin Y, Skaggs W E, Suster M, Weaver K L 1995 Deciphering the hippocampal polyglot: The hippocampus as a path integration system. Journal of Experimental Biology 199: 173–85
16. Muller R U, Kubie J L 1987 The effects of changes in the environment on the spatial firing of hippocampal complexspike cells. Journal of Neuroscience 7: 1951–68
17. Muller R U, Stead M, Pach J 1996 The hippocampus as a cognitive graph. Journal of General Physiology 107: 663–94
18. O’Keefe J, Burgess N 1996 Geometric determinants of the place fields of hippocampal neurons. Nature 381: 425–28
19. O’Keefe J, Conway D H 1978 Hippocampal place units in the freely moving rat: Why they fire where they fire. Experimental Brain Research 31: 573–90
20. O’Keefe J, Dostrovsky J 1971 The hippocampus as a spatial map: Preliminary evidence from unit activity in the freely moving rat. Brain Research 34: 171–75
21. O’Keefe J, Nadel L 1978 The Hippocampus as a Cognitive Map. Oxford University Press, Oxford, UK
22. Ranck J B Jr 1984 Head-direction cells in the deep cell layers of the dorsal presubiculum in freely moving rats. Society of Neuroscience Abstracts 10: 599
23. Redish A D, Elga A N, Touretzky D S 1996 A coupled attractor model of the rodent head direction system. Network 7: 671–86
24. Samsonovich A, McNaughton B L 1997 Path integration and cognitive mapping in a continuous attractor neural network model. Journal of Neuroscience 17: 5900–20
25. Sharp P E 1991 Computer simulation of hippocampal place cells. Psychobiology 19: 103–15
26. Sharp P E, Blair H T, Brown M 1996 Neural network modeling of the hippocampal formation spatial signals and their possible role in navigation: A modular approach. Hippo- campus 6: 720–34
27. Sharp P E, Green C 1994 Spatial correlates of firing patterns of single cells in the subiculum of the freely moving rat. Journal of Neuroscience 14: 2339–56
28. Skaggs W E, Knierim J J, Kudrimoti H S, McNaughton B L 1995 A model of the neural basis of the rat’s sense of direction. In: Tesauro G, Touretzky D S, Lean T K (eds.) Advances in Neural Information Processing Systems 7. MIT Press, Cambridge, MA
29. Takahashi N, Kawamura M, Shiota J, Kasahata N, Hirayama K 1997 Pure topographic disorientation due to right retrosplenial lesion. Neurology 49: 464–69
30. Taube J S 1995 Place cells recorded in the parasubiculum in freely moving rats. Hippocampus 5: 569–83
31. Taube J S, Muller R U, Ranck J B Jr 1990a Head direction cells recorded from the postsubiculum in freely moving rats. I. Description and quantitative analysis. Journal of Neuroscience 10: 420–35
32. Taube J S, Muller R U, Ranck J B Jr 1990b Head direction cells recorded from the postsubiculum in freely moving rats. II. Effects of environmental manipulations. Journal of Neuroscience 10: 436–47
33. Teng E, Squire L R 1999 Memory for places learned long ago is intact after hippocampal damage. Nature 400: 675–77
34. Tolman E C 1948 Cognitive maps in rats and men. Psychological Review 55: 189–208
35. Treves A, Miglino O, Parisi D 1992 Rats, nets, maps, and the emergence of place cells. Psychobiology 20: 1–8
36. Zhang K 1996 Representation of spatial orientation by the intrinsic dynamics of the head direction cell ensemble: A theory. Journal of Neuroscience 16: 2112–26
37. Zipser D 1985 A computational model of hippocampal place fields. Behavioral Neuroscience 99: 1006–18