The ability to simultaneously monitor hundreds to thousands of neurons in the brain is enabling an experimental shift of attention from the activity of single cells (which, until recently, was usually measured as a function of external correlates), to the internal dynamics of large neural populations. One of the most striking outcomes of this recent development is observed in brain areas that represent an animal’s position in its environment, where neural populations have been observed to robustly express activity patterns that reside on low-dimensional nonlinear manifolds, in agreement with theories that were put forth in the past several decades. In the grid cell system, found in the entorhinal cortex of mammals, single cells are active in multiple spatial locations that are arranged periodically on a hexagonal lattice. Cells are functionally arranged in so-called modules: within a module, grid cells share the same spacing of their periodic spatial responses.  Very recently, it was shown that grid cells within a module robustly express activity patterns that reside on a two-dimensional manifold with toroidal topology. I will focus on the joint dynamics of neural activity in multiple grid cell modules: we recently found that different modules tightly coordinate their activity even when the brain’s internal representation of position is dissociated from the true position of the animal. I will present the evidence for this coordination, based on simultaneous recordings of hundreds of grid cells in animals that were foraging in the dark. I will also discuss theories for how this coordination might be implemented by neural circuitry in the brain.