Eppendorf and Science Prize

The Grid Map in the Brain

Marianne Hafting Fyhn

Most of the time we are not aware of it, but the brain constantly keeps track of where we are. In order to get through our daily lives, we depend upon the brain's ability to keep track of position, direction, and distance. The neural mechanisms underlying such spatial navigation have long been an area of intense study and lively debate in neuroscience. A milestone was achieved when O'Keefe and Dostrovsky (1) discovered that neurons in the dorsal hippocampus were active only when the test animal (a rat) was in a specific place. Such "place cells" have been the subject of more than 30 years of research pointing to the hippocampus as a key component of the brain's circuit for spatial computations (2). However, a growing number of studies have shown that place cells are influenced both by changes in the environment and by previous experiment (3-9), and thus these cells do not code solely for location. Furthermore, disconnecting the hippocampal subareas suggested that the origin of the spatial code was not in the hippocampus, but rather in some of the input areas (10). A likely candidate was the entorhinal cortex, which is situated only one synapse upstream of the hippocampus and receives sensory information of all modalities (11). Previous studies found only weak spatial modulation of entorhinal neurons (12, 13), but these recordings were not from the region of entorhinal cortex where spatial computations are most likely to occur. The entorhinal cortex shows a modular organization in which the dorsocaudal part of medial entorhinal cortex (MEC) receives most of the visual and spatial information and in turn projects to the dorsal hippocampus (14) where the sharp and distinct place fields are found. To test the hypothesis that spatial calculations are performed upstream of the hippocampus, I recorded extracellular spike activity of single neurons from the entire dorsoventral axis of MEC in free-roaming rats chasing biscuit crumbs in a 1 m by 1 m enclosure (15).

We found strong spatial modulation of neurons in dorsal MEC and weaker spatial modulation more ventrally, corresponding to the anatomical modular organization of MEC (16). The dorsocaudal neurons had multiple firing fields, and the spacing was highly regular. Therefore, to capture any regularity in the spatial firing structure, we doubled the size of the recording arena. The firing fields of each cell portrayed a remarkable hexagonal pattern of regular triangles covering the entirety of the animal's environment (see the figure) (17). We termed them "grid cells." The pattern resembled the cross-points of a sheet of graph paper rolled out over the surface of the test arena, but with equilateral triangles rather than squares as the repeating unit. Such a pattern of symmetric receptive fields could not result from external sensory input alone but must also be due to pattern generation within the brain itself. Furthermore, the activity of eight grid cells was enough to reconstruct the running path of the rat. In repeated exposures to the same environment, the grid cells fired at the identical positions, suggesting that grid cells constitute a stable map of the environment and that they are part of a neuronal coordinate system for spatial navigation. In contrast to the hippocampal map, the grid map was activated in a stereotypic manner across environments, regardless of the environment's particular landmarks, suggesting that the same neural map was applied everywhere. However, the grid seems to be driven by self-motion cues because it forms instantaneously in a novel environment and is not perturbed by removal of visual cues. Because of its gridlike nature, the grid map is potentially infinite and may represent places not visited as well as places that have been visited. The existence of a single neural map that can be applied anywhere is efficient and avoids the capacity problem of having separate maps for every spatial context.

In addition to information about position and distance encoded by the grid cells, directional information is needed for proper navigation. The different cell layers of MEC receive directional input from neurons, whose activity is correlated to the head-direction of the animal ("head-direction cells") (18), and context-specific place-information from the hippocampus. To investigate how the navigational information is integrated in the entorhinal network, we recorded from all layers of MEC (19). Layer II was predominated by grid cells, which were colocalized with head-direction cells and cells with mixed grid- and head-direction properties in deeper layers. These results point to the integration of positional, directional, and translational information in single MEC cells. The anatomical position and functional properties of MEC suggest that MEC is the hub of a distributed brain network for navigation.

Spatial information is not used exclusively for navigation but constitutes a framework for autobiographical memories. One can recall similar events that happened at the same location without mixing up the memories. The population dynamics of grid cells and their relation to memory representations in the hippocampus were unknown. To address this issue, we recorded simultaneously from grid cells and hippocampal place cells in environments with different degree of similarity (20). We showed that the disambiguation of representations in the hippocampus is determined by ensemble dynamics in entorhinal grid cells. In environments with pronounced differences, the hippocampus recruits different cell ensembles (8). This "remapping" was invariably preceded by a coherent coordinate shift in the ensemble map in the entorhinal cortex. However, small changes in the environment were only detected by the hippocampal ensemble, confirming that the hippocampal formation is responsible for pattern separation (21-23).

It remains unclear whether humans have grid cells. Damage to the medial temporal lobe, which includes the hippocampus and entorhinal cortex, causes severe memory loss in mammals (24). During early stages of Alzheimer's disease, symptoms include impairments in the sense of locality (25), and the first area to exhibit pathological cell loss is the entorhinal cortex (26, 27). Thus, it is likely that MEC plays a pivotal role in human navigation.

The discovery of grid cells paves the way for a new and mechanistic understanding of navigation and of memory formation.


  1. J. O'Keefe, J. Dostrovsky, Brain Res.34, 171 (1971).
  2. J. O'Keefe, L. Nadel, The Hippocampus as a Cognitive Map (Clarendon, Oxford, 1978).
  3. E. Bostock, R. U. Muller, J. L. Kubie, Hippocampus1, 193 (1991).
  4. E. J. Markus, Y. L. Qin, B. Leonard, W. E Skaggs, B. L. McNaughton, C. A. Barnes, J. Neurosci.15, 7079 (1995).
  5. S. A. Hollup et al., J. Neurosci.21, 1635 (2001).
  6. E. R. Wood et al., Neuron27, 623 (2000).
  7. M. Fyhn, S. Molden, S. Hollup, M.-B. Moser, E. I. Moser, Neuron35, 555 (2002).
  8. S. Leutgeb et al., Science305, 1295 (2004).
  9. M. A. Moita et al., J. Neurosci.24, 7015 (2004).
  10. V. H. Brun et al., Science296, 2243 (2002).
  11. R. D. Burwell, Ann. N. Y. Acad. Sci.911, 25 (2000).
  12. G. J. Quirk et al., J. Neurosci.12, 1945 (1992).
  13. L. M. Frank, E. N. Brown, M. A. Wilson, Neuron27, 169 (2000).
  14. C. L. Dolorfo, D. G. Amaral, J. Comp. Neurol.398, 25 (1998).
  15. M. Fyhn, S. Molden, M. P. Witter, M.-B. Moser, E. I. Moser, Science305, 1258 (2004).
  16. C. L. Dolorfo, D. G. Amaral, J. Comp. Neurol.398, 49 (1998).
  17. T. Hafting, M. Fyhn, S. Molden, M.-B. Moser, E. I Moser, Nature436, 801 (2005).
  18. J. S. Taube, Prog. Neurobiol.55, 225 (1998).
  19. F. Sargolini et al., Science312, 758 (2006).
  20. M. Fyhn, T. Hafting, A. Treves, E. I. Moser, M.-B. Moser, Nature 446,190 (2007).
  21. E. T. Rolls, A. Treves, Neural Networks and Brain Function (Oxford Univ. Press, Oxford, 1998).
  22. J. K. Leutgeb et al., Neuron20, 345 (2005).
  23. J. K. Leutgeb et al., Science315, 961(2007).
  24. M. Suzuki, H. Eichenbaum, Ann. N. Y. Acad. Sci.911, 175 (2000).
  25. V. Kavcic et al., Brain129, 736 (2006).
  26. C. F. Lippa et al., Neurobiol. Aging3, 405 (1992).
  27. T. Gomez-Isla et al., J. Neurosci.16, 4491 (1996).
  28. I thank my colleagues at the Centre for the Biology of Memory for their collaboration in this research, and especially my supervisors M.-Britt and E. Moser and my husband T. Hafting. This research was supported by a Research of Excellence Grant from the Norwegian Research Council.