
Using Grid Cells for Navigation


The regular and repeated triangular spatial firing pattern of entorhinal grid cells suggests a strong pathintegration (idiothetic) input. At the same time, trialtotrial repeatability and lack of intratrial drift suggest locationspecific sensory drive. Our basic model of hippocampal circuitry involves a reciprocal relation between hippocampal place cells and grid cells: grid cells provide idiothetic input to place cells and place cells provide locationspecific sensory input to grid cells.
Our model for the contribution of grid cells to navigation follows from our earlier work on headingvector navigation between pairs of familiar locations (landmarks). Here we are suggesting that hippocampal place cells identify startpoints and endpoints with unique realworld locations, while the gridcell universal map provides the connecting heading vector. More specifically, startpoint place cells point to a firing vector of grid cells (the startpoint gridcell set). Similarly, goal place cells point to a second firing vector of grid cells (the goal gridcell set). Each gridcell set includes cells of numerous scales, but, likely, a single orientation. We are interested in three questions: (1) knowing the startpoint and goal gridcell sets, can the heading vector connecting these be computed? (2) How can the computation be performed? (3) Is this computation implemented in the rat brain?
Potential models for headingvector computation depend heavily on the organizational structure of entorhinal cortex. Specifically, different types of models apply if the DV distribution of grid scale is steplike or continuous. Additionally, one class of models depends on grid cells of a fixed scale having highly constrained scale and “rigid” spatial relations between pairs of cells across all locations and all environments. These models will be described and explored. We have begun recording grid cells in a large rectangular enclosure (1.8m x 1.4m) using a 3D tracking system to eliminate optical and parallax distortion. Preliminary findings from 8 cells recorded from one rat support the discretestep organizational pattern of grid cell scale: 7 of 8 cells had virtually the same scale, while the 8th was larger by a factor of 1.45. Additionally, all 8 cells exhibited virtually identical grid orientations. These findings suggest that each set of grid cells with identical scale form a rigid network in that movement of any direction and distance from one firing vector of grid cells will predict a unique firing vector.