Coding for spatial navigation using grid cells


March 17, 2016 - 1:00pm
Northwest 243
About the Speaker
Martin Stemmler (LMU)

Mammalian grid cells fire when an animal crosses the points of an imaginary, hexagonal grid tessellating the environment. Is the grid-like structure of firing fields a consequence of a continuous attractor network within medial entorhinal cortex? I will show evidence against this hypothesis and argue that neural coding constraints explain the geometry of the spatial code. In this talk, we will explore how animals can navigate by reading-out a simple population vector of grid cell activity across multiple scales. Dead reckoning in navigation requires grid cells to be organized into discrete modules, such that each cell within a module has the same lattice scale and orientation. Across modules, the spatial periods should change, forming a geometric progression with a scale ratio of around 3/2. This ratio minimizes the risk of making large-scale errors in navigation. Silencing intermediate-scale modules should cause systematic errors in homing, while knocking out the module at the smallest scale will only affect navigational precision. The read-out neurons behave like goal-vector cells in the population vector framework, transforming the world-centered coordinates inherent in the grid cells' periodic spatial tuning into the body-centered coordinates needed for planning movements. Nonlinear gain fields set the goal locations. We propose that the read-out neurons might be found outside of the hippocampus proper, for instance, in retrosplenial or posterior parietal cortex.