CNS Seminar

In this talk I'll explore the striking tuning properties of grid cells from a theoretical perspective, with the aim of shedding some light on both their function and mechanism.

First, I'll discuss why the brain might choose to represent location, a non-periodic, local variable, using grid cell spatial tuning curves, which are periodic and non-local. I'll show that in principle, the grid cell system is capable of exponential representational capacity using linearly many neurons, in contrast to other well-characterized neural codes, which tend to exhibit a linear increase in capacity with neuron number.

Next I'll turn to the question of circuit architecture and dynamics. I'll sketch candidate network models of grid cell networks, then show  that a detailed analysis of pairwise grid cell data reveals a very low-dimensional population response. This finding tilts the evidence strongly in favor of one group of models. I'll give a developmental account of how such network structures can develop, and end with a theoretically motivated proposal for simple experimental perturbations that should reveal in much richer detail the neural circuit mechanisms that underlie the remarkable grid-cell code for location.