Abrupt resetting of the pattern occurred at the turning point

between each corridor, suggesting that salient landmarks or environmental features may reset the periodic pattern, resulting in a fragmented map. Based on the ability of the grid pattern to fragment in a complex environment, the encoding capacity of the grid network might also increase by representing environments as mosaics of smaller spatial maps (Derdikman and Moser, 2010 and Derdikman et al., 2009). In large environments, smaller maps may split along salient environmental borders or features. Readout of location from multiple map fragments may then rely on a mechanism or Angiogenesis inhibitor brain system substantially different from the metric readout of grid and place cells. Finally, a functional understanding of grid cells would need to incorporate the fact that natural environments have a three-dimensional topography that is very different from the flat, unconstrained surfaces that rodents explore in the laboratory. A recent paper probed how grid cells map to the vertical dimension by requiring rats to explore a helix-shaped track as well as a vertical surface lined with protruding horizontal pegs (Hayman et al., 2011). In these environments, grid fields

appeared as vertical columns, suggesting that grid cells do not differentiate between x,y positions with different z coordinates. However, it cannot presently be ruled out that periodic fields would reappear if animals are allowed to move continuously in the vertical plane, in the same way that they move Doxorubicin on horizontal surfaces. Another possibility is that the scale of grid cells is larger in the vertical plane and that the column structure simply reflects a stretched grid with a significantly larger field GBA3 size. Testing these possibilities is challenging but may be possible in species other than rats and mice. Our understanding of spatial representation in the hippocampal-entorhinal system has been strongly influenced by computational models. Models have proposed possible mechanisms for formation and transformation of

spatial firing patterns, and they have constrained the ways in which such patterns can be generated in circuits with known properties. Each model that we have described makes a number of testable predictions, but verification and falsification have so far remained indirect in that the experimental evidence is mostly correlational and subject to multiple interpretations. With the development of more sophisticated experimental methods during the next few years, the interaction between theory and experiment will likely be strengthened. Direct testing of models of grid cells may require quantitative analysis of the intrinsic dynamics and connectivity of individual neurons, and it may be necessary to activate as well as inactivate specific inputs to these neurons.