Event:

Grid cells in rodents and bats provide a glimpse into the coding, deep within the brain, of an internally computed quantity - an animal’s self estimate of its position in its environment. Recent experiments established that these cells are functionally organized in discrete modules, each containing neurons with uniform grid spacing, whereas the spacings seen in successive modules approximately follow a geometric progression. This result is in agreement with theories that postulate a role of grid cells in efficient coding of the animal’s position. However, the experimental data suggests also that the number of cells decreases sharply with grid spacing, in marked disagreement with existing theories. I will introduce a hypothesis that the entorhinal cortex is adapted to represent a dynamic quantity (the trajectory of the animal in space), while taking into account the temporal statistics of this variable. We recently developed a theory for efficient coding of such a variable. A central prediction of the theory is that neuron population sizes should sharply decrease with the increase of grid spacing, in agreement with the trends seen in the experimental data. I will also discuss a simple, near optimal scheme for neural readout of the dynamic position from the grid cell code, in which model place cells linearly sum inputs from grid cells. Crucially, the summation involves a temporal kernel, whose characteristic decay time depends on the spacing of the presynaptic grid cell. The simple readout scheme requires mechanisms for persistence over time scales ranging from 1 ms to 1 s, suggesting that diverse biophysical mechanisms for persistence may be involved in readout of the grid cell code.