The lattice scales of grid-cell activity

Grid cells in the rodent medial entorhinal cortex (MEC) form a multi-scale representation of the animals environment and are thought to be a signature of continuous-attractor networks. Experiments suggest that grid scales are discrete and arranged in a geometric progression with constant scale ratios, as predicted by efficient-coding theories that posit that the grid-cell system is optimized for spatial resolution. This normative explanation has been challenged by a developmental theory whereby grid-cell modularity arises from pattern formation in a system with smooth parameter gradients. As a consequence, scale ratios should not be constant, but vary by module. However, as we show here, the experimental data chosen to support this mechanistic theory fail to do so. Publicly available large-scale grid-cell datasets could provide a clear benchmark and further our understanding of the neuronal basis of spatial navigation. We, therefore, reanalyzed such recordings and found that they refute the developmental theory. Instead, the measured scale ratios agree with the geometric-progression hypothesis and set strict bounds for any future grid-cell theory.